![Generating Strongly Connected Random Graphs
Peter M. Maurer
Dept. of Computer Science
Baylor University
Waco, Texas 76798
Abstract – The algorithm presented here is capable of generating
strongly connected graphs in a single pass that requires O(n) time.
The method is to create a spanning tree that is a directed acyclic
graph, and adding a minimal number of edges to make the
spanning tree strongly connected. This is done in a way that is
completely general. Once the strongly connected spanning tree is
created, additional edges can be added to the tree to create an
arbitrary strongly connected graph.
1 Introduction
One important problem in many types of simulation is creating
random data for input to the simulation. Over the years, our need for
such data in the simulation of gate-level circuits has led us to create a
package for creating many different types of random data [1,2].
Despite its utility, this package has become dated. It was originally
intended to generate random data files for input to a simulation
program. Recently, we have begun a project to upgrade this package
with new features to make it more useful for modern types of programs
that do not depend heavily on file-based input, and to generate types
of data that are more suitable to modern programs than character
strings and simple binary values.
One major focus of this activity (there are many) is the generation
of random graphs. Graphs can be used to model many real-world
phenomena. There are too many applications to mention individually,
but see [3] for an example. One new feature of our package is the
creation of graph-generation subroutines that can be incorporated into
existing software. The graphs are generated internally as adjacency
lists and passed, as pointers, to the simulation software.
Graph specifications are simple, typically one line, but permit the
specification of many different types of random graphs. The most
common models are the edged-oriented models, the Gilbert model [4]
and the Erdős–Rényi model [5]. The vertex-oriented models, power-
law [6] and degree-sequence [7] models are also fairly common. This
paper will focus on the edge-oriented models. The Gilbert model
assigns a probability of p to the existence of any edge, and the Erdős–
Rényi model assigns equal probability to all graphs with M edges.
For the Gilbert model we generate k vertices and add each edge from
the complete graph with probability p . For the Erdős–Rényi model,
we sort all edges into random order and choose the first M edges from
the sorted list. (The parameters k, i, and M are specified by the user.)
The power-law and degree-sequence models are also available in our
package, but these are beyond the scope of this paper.
Parameters can be used to specify that the graph is directed, or
that the graph must be connected, or both. Creating a connected non-
directed graph is relatively simple. We generate a spanning tree using
Algorithm 1, and then apply either the Gilbert or the Erdős–Rényi
model to the remaining edges. The purpose of Algorithm 1 is to create
a spanning tree for the graph. A non-directed graph is connected if and
only if a spanning tree exists. Algorithm 1 adds Vertex 0 to the
spanning tree, and then adds the remaining vertices by selecting a
random vertex from the partial spanning tree as the parent of the new
vertex.
1. Add Vertex 0 to the tree.
2. For each vertex, i, 1 through 1
k
a. select a vertex j at random from the existing tree
vertices
b. Add an edge between i and j.
c. Add vertex i to the tree.
Algorithm 1. Creating the Spanning Tree.
When the graph is directed, simply creating a spanning tree is
insufficient because the resultant graph must be strongly connected.
The spanning tree is still necessary, because there must be a path from
Vertex 0 to every other vertex. When creating a spanning tree for a
directed graph, the first step is to modify step 2 of Algorithm 1 so that
the new edge proceeds from the tree vertex to the new vertex. This
insures that there is a path from Vertex 0 to every other vertex. The
resultant tree is a directed acyclic graph with the root of the tree as the
only source. The sinks are the leaves of the tree.
There are several straightforward methods for making the
spanning tree strongly connected. When adding a tree vertex, we could
add two edges, one from the tree vertex to the new vertex, and another
in the opposite direction. This would make the tree strongly connected,
but there would be no way to generate certain types of graphs such as
the simple cycle of Figure 1. Another method is to add an edge from
each leaf vertex to the root vertex. This is perhaps more general, but
graphs such as that pictured in Figure 2 would be impossible to
generate.
Figure 1. A Simple Cycle.
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 3
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-15-320.jpg)
![Figure 2. Impossible for Leaf-to-Root.
It is necessary to be able to generate any strongly connected graph
on k vertices. Some attempts have been made to do so using the
rejection method [8]. (The rejection method repeatedly generates
directed graphs at random, and rejects those that are not strongly
connected.) However, the rejection method works well only if the
probability of rejection is small. When a sparse strongly connected
graph is required, the rejection method is not suitable.
Creation of the spanning tree is clearly the starting point for such
a procedure, since it must exist. Furthermore, Algorithm 1 is able to
generate any spanning tree, up to isomorphism. The next step must be
to add additional edges to the graph to make it strongly connected.
Furthermore, this should be done in a way that is certain to succeed
and adds only a minimal number of edges to the tree. (The Leaf-to-
Root method generates the absolute minimum number of edges, but is
not suitable because it is not able to generate all strongly connected
graphs.)
To meet the needs of our generation software, we use a modified
version of the Tarjan strongly connected component algorithm [9]. The
Tarjan algorithm is capable of identifying any strongly connected
graph by performing a single depth first search. The forward edges of
the depth first search define a tree which is equivalent to our initial
spanning tree. To detect a strongly connected graph, the Tarjan
algorithm identifies a minimal set of back and cross edges to insure
that the initial spanning tree strongly connected. Rather than detecting
such edges (which do not exist in the initial spanning tree) we modify
the algorithm to insert such edges where required. We do this in such
a way that any suitable set of back or cross edge can be generated.
Once the tree has been made strongly connected the Gilbert or Erdős–
Rényi model can be applied to the remaining edges.
2 The Tarjan Algorithm
To understand our method of inserting edges into the tree it is
necessary to understand the principles of the Tarjan algorithm. The
mechanism is based on depth first search with computed start times.
Algorithm 2 shows the basic depth first search algorithm with the
modification points that are used to detect strongly connected
components. Start time is an integer in the range [0, 1]
n where n is
the number of vertices in the graph. The start time of vertex i is a
sequential number indicating the order in which vertex i was first seen
by the depth first search algorithm. In Algorithm 2, each element of
the StartTime array (which is of size n) is initialized to 1
, and
GlobalStartTime is initialized to zero. Both are global variables.
1. Function call DFS(i)
2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen
3. Increment GlobalStartTime by 1.
4. For each vertex
j adjacent to vertex i
a. If StartTime[j] is equal to 1
Call DFS(j) // back
up to i
5. // backing up from i
Algorithm 2, basic DFS.
To detect strongly components, the basic DFS algorithm must be
modified in three places. First, we add an array of size n named Low.
After a vertex has been completely processed, it Low[i] will contain
the smallest start time of any vertex that can be reached by following
zero or more tree edges from Vertex i, followed by one back edge or
cross edge. The following is added after step 2.
2.5. Set Low[i] equal to StartTime[i].
This step indicates that, initially, the lowest reachable start time
is the start time of the current vertex. To step 4a we add Low[i] =
min(Low[i],Low[j]). If it is possible to get further back than the current
value of Low[i] by using tree edges, the new minimal start time
recorded.
The following step is added after step 4a:
b. else if j is not already in a SCC, Low[i] =
min(Low[i],StartTime[j])
If the edge (i,j) is a cross edge or a back edge, and it is further
back than we have been able to get previously, its start time is
recorded. The only problem that arises is when (i,j) is a cross edge to
a strongly connected component that has already been identified. Such
edges must be ignored. A status-array is normally used to keep track
of such vertices.
Finally, Step 5 is added to identify strongly connected
components. If Low[i] is equal to StartTime[i], then there is no back
edge or cross edge that provides a path from Vertex i to its parent.
Therefore, Vertex i is in a different strongly connected component than
its parent.
5. If Low[i] is equal to StartTime[i] Identify a new SCC.
Tarjan’s algorithm also includes a stacking mechanism to identify
the vertices belonging to a particular strongly connected component,
but because our aim is to create a single strongly connected
component, we will not consider this mechanism further. The same is
true for the mechanism that tags vertices that have already been
assigned to a strongly connected component. The full Tarjan algorithm
is given in Algorithm 3.
4 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-16-320.jpg)
![1. Function call DFS(i)
2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen
3. Set Low[i] equal to StartTime[i]
4. Increment GlobalStartTime by 1.
5. For each vertex
j adjacent to vertex i
a. If StartTime[j] is equal to 1
Call DFS(j), Low[i]
= min(Low[i],Low[j])
b. Else if
j is not already in a SCC, Low[i] =
min(Low[i],StartTime[j])
6. If Low[i] equals StartTime[i] Identify a new SCC
Algorithm 3. The Full Tarjan Algorithm.
3 The modified Tarjan algorithm
Our primary modification to the Tarjan algorithm is in Step 6,
which identifies new strongly connected components. We wish to
prevent this detection from taking place. In Step 6, the algorithm is
backing up from Vertex i . All vertices with start times greater than or
equal to StartTime[ i ] are in the subtree rooted at Vertex i . All
vertices, j , that have start times less than i must meet one of the
following three conditions.
1. Vertex j is an ancestor of Vertex i in the DFS tree.
2. Vertex j has already been assigned to a strongly connected
component.
3. Vertex j and Vertex i have a common ancestor k , and k
is reachable from j .
Ignoring condition 2, it is clear that in Step 6, that Vertex i will
be in the same strongly connected component as its parent if and only
if there is an edge from a vertex v with a start time greater than or
equal to StartTime[ i ] to a vertex w with a start time less than
StartTime[ i ]. We modify Step 6 to insert such an edge when Low[i] is
equal to StartTime[i]. Doing this also insures that condition 2 can never
occur. Our modified algorithm is given in Algorithm 4.
1. Function call DFS(i)
2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen
3. Set Low[i] equal to StartTime[i]
4. Increment GlobalStartTime by 1.
5. For each vertex
j adjacent to vertex i
a. If StartTime[j] is equal to 1
Call DFS(j), Low[i]
= min(Low[i],Low[j])
b. Low[i] = min(Low[i],StartTime[j])
6. If Low[i] equals StartTime[i]
a. Set x to a random integer in the range
[StartTime[i],GlobalStartTime]
b. Set y to a random integer in the range
[0,StartTime[i]-1]
c. Identify the vertices v and w that correspond to the
start times x and y.
d. Add an edge from v to w.
e. Set Low[i] equal to y.
Algorithm 4. The Modified Tarjan Algorithm.
Step 6c requires the inversion of the function that assigns start
times to vertices. This is done in constant time by using an Inverse
Start Time (IST) array and the following step following Step 2.
2.5 Set IST[StartTime[i]] to i
In Step 6c, v and w correspond to IST[x] and IST[y], respectively.
Step 6e negates the Low[i] equals StartTime[i] condition, since Vertex
i is now in the same strongly connected component as its parent. The
addition of the new edge may make the Low value for some of the
other vertices in the tree rooted at Vertex i incorrect, but since these
values will not be accessed after backing up from Vertex i, they do not
need to be changed.
Once Algorithm 4 has been run, the Gilbert or Erdős–Rényi
models can be applied to add additional edges to the graph. The graphs
created by Algorithms 1 and 4 are strongly connected and contain no
more than 2 2
n edges. Algorithm 1 adds 1
n edges. Algorithm 4
can add at most one edge per vertex, and cannot add an edge to Vertex
0. Algorithm 4 can add edges in any possible way to make the tree
strongly connected. When combined with the Gilbert or Erdős–Rényi
models any strongly connected graph can be generated. Both
Algorithm 1 and Algorithm 4 are O(n). The Gilbert or Erdős–Rényi
models are both worst-case O(
2
n ) because they must consider all
2
2
n n
potential edges.
Although Algorithms 1 and 4 can generate any strongly
connected tree up to isomorphism, there are many isomorphs that will
never be generated. For example, Algorithm 1 always adds an edge
from Vertex 0 to Vertex 1. If this is a problem, it can be solved by a
random relabeling of the vertices after the graph is generated.
4 Experimental Results
We ran several experiments to verify the effectiveness of our
algorithm. The first experiment was to generate 10,000 5-vertex
graphs, using the Gilbert model with 0
p , to verify that the examples
of Figures 1 and 2 could be generated. The algorithm generated both
these examples, along with many others. Figure 3 contains a sample of
the generated graphs.
Figure 3. Some Sample Graphs.
Figure 3 demonstrates the essentially random nature of the
generation process. Despite the fact that there is always an edge
between Vertex 0 and Vertex 1, the structure of the graphs is obviously
quite random, and they are obviously all strongly connected. This
experiment was run using the Gilbert model with edge probability set
to zero. This was done to show the structure of the strongly connected
tree.
Four other experiments were run to determine the performance of
the algorithm. The hardware was an Intel 3.40 Ghz core I7-2600 with
4 cores and 8GB of memory, running Linux Red Hat version 3.10. A
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 5
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-17-320.jpg)
![Simulating Virtual Memory Allocations Using SPEC Tools in
Microsoft Hyper-V Clouds
John M. Medellin*, Lokesh Budhi
Associate Professor and Graduate Assistant at the Master of Science in Information Systems
University of Mary-Hardin Baylor
Belton, TX 76513-2599, USA
Abstract— Private Clouds are gaining in popularity with
small to medium sized businesses. By implementing a
virtualized architecture, a company can gain strategic
advantage through higher utilization of their technology
assets. One of the first steps in determining if the virtualized
architecture will make sense is to estimate the amount of
resources required to actually create a private cloud.
A risky approach is to take the applications that are
executing and try to run them on a cloud. This option could
turn out to be costly since key legacy applications are tightly
coupled and in order to run the experiment one might need
to move the entire system over (build the entire Cloud and
port them). A second approach could be to model key parts
of the system and test them with empirical models. This
could also be costly and risky if key characteristics are
erroneously estimated or omitted. Perhaps a better
approach could be to use an industry simulation that can
predict the usage patterns of similar systems and be
configured to resemble workloads in production today.
This paper executes simulations both on bare
metal and within the Microsoft Cloud Stack (Windows 10,
Windows Server 2012 R2 and Windows Hyper-V 2016)
using the industry standard SPECjbb2015 simulation
environment. We focus on measurement of incremental
memory allocation and report throughput differences from
two bare metal architectures (Windows 10 and Windows
Server 2012 R2) to the target private cloud architecture.
Our work begins by allocating 8GB to each environment
and increases that variable to 10GB and 12GB. Significant
performance gains are gained by increasing memory
allocation in the virtual machine.
We believe the contribution of this work is to
demonstrate how industrial strength simulation tools can be
applied to real world scenarios without having to completely
build-out the architectures considered. This should be
particularly useful to small companies that are
contemplating private cloud implementations.
Keywords— Hypervisors, Workload Simulation,
Retail Applications, SPEC Corporation, Microsoft
Windows 10, Microsoft Windows Server 2012 R2,
Microsoft Hyper-V 2016
I. INTRODUCTION
Clouds are used by many people and organizations today
to gain a variety of advantages. There are many vendors and
open sources for cloud software and an equal number of
techniques for evaluating them. We can mix and match
products that take advantage of our particular situation and
the application workload profiles we are targeting. Each
candidate architecture performs and enhances certain types of
applications (referred to as “workloads”). The final selection
will probably be based on the types of projected applications
and their workload profiles [10]. Once we select the target
Cloud tools they will also have to be tuned as far as memory
allocations to deliver the expected results.
Many studies have been published on the impact of
resource virtualization and workload characteristics on Cloud
architectures. Clouds essentially contain Virtual Machines
and are managed by a central authority called a Virtual
Machine Manager or a “Hypervisor”. Hypervisors can be
secured from traditional vendors (e.g., Microsoft Hyper-V)
or on open source like the OpenStack project [9]. Hypervisors
are configured to either interact with the hardware directly
(Type 1) or through a Host Operating system (Type 2) [3]. A
typical implementation in smaller installations is the
Microsoft Hyper-V, a type 2 architecture that runs on top of
Windows Server 2012 R2 and manages virtual machines that
can have Windows 10 or other guest operating systems. The
applications themselves execute inside the virtual machines
on the guest operating systems.
A key strategy for measuring the performance of certain
architecture attributes is the selection of an industry-standard
simulation tool that will lend credibility to the results (e.g. it
resembles what is to be measured). Simulation suites for
measurement of a variety of attributes are provided by the
Standards Performance Evaluation Corporation “SPEC”;
www.spec.org. SPEC is a non-profit organization that was
created by a consortia of major technology providers who
have agreed on a set of principles to be used in building
benchmarking tools. SPECjbb2015 is a simulated transaction
generator that can provide for very complex scenarios in a
retail grocery store. The system provides a set of simulation
tools that can be applied to build a scalable model to resemble
reality. When used to simulate Cloud performance, the tool
will deliver compute transaction workloads (impacts on the
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 7
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-19-320.jpg)
![CPU/memory) that are executed in bare metal or on Virtual
Machines (VMs). The functionality injects a series of
standard transactions into a set of application processes that
drive simulation through java applications provided. These
transactions are progressively injected into the applications
until the system gets saturated and can no longer provide
sufficient throughput to keep up with the load being input
(inputs exceed outputs).
The fundamental objectives of this research are to first
simulate the incremental overhead added by virtualization
from bare metal to the Hyper-V environment and second to
add more memory to determine the effects on overcoming the
virtualization penalty. We use the SPECjbb2015 to simulate
on bare metal using the Windows 10 operating system and
the Windows Server 2012 R2 allocating 8GB in each case.
Next, we virtualize that environment in a VM on Hyper-V,
allocate the same amount of memory and execute the
simulation. In a second experiment, we increase the allocated
memory to 10GB and 12 GB and report the throughput
statistics. The business applicability of the approach is
discussed at the end of the document.
In our analysis, we first present related work that has been
done and how we have adapted some of those methods/results
in our research. Next, we create a series of experiments which
simulate the impact of virtualization as follows:
• The SPECjbb2015 is executed on bare metal under the
Windows 10 and the Windows Server 2012 R2
Operating Systems. Next, we virtualize the simulation
under Windows Server 2012 R2 Network OS with the
Hyper-V 2016 Hypervisor and Windows 10 guest
operating system. All three of these have 8GB of
Memory allocated. The results are reported in
SPECjbb2015 throughput transaction totals.
• The simulations above are repeated except under
varied memory allocation at 10GB and 12GB. The
corresponding increase throughput totals is reported.
This research aims to demonstrate the usage of standard
simulation tools in order to determine potential alternatives
in Cloud resources without having to build the specific
environments. The approach used could be scaled to other
Cloud architectures than the one presented.
II. RELATED WORK
Virtualized environments date back a few decades. A key
objective of virtualization was to keep the CPU busy while
memory variables were being fetched from slower
components in the computer [12]. With the advent of fully
logically defined architectures in software (“software-
defined systems”) we are now able to abstract the physical
components into specifications resident in configuration
files. The key software agent that manages and provisions the
resources in a modern cloud is the Virtual Machine Monitor
also referred to as the “hypervisor” [3]. All policies regarding
allocation and usage of physical infrastructure resources are
controlled by the hypervisor. Hypervisors are assisted by
other tools and agents in order to deliver a fully functional
Cloud Management Platform (CMP) [5].
A. Hypervisor Architecture Throughput
In their review of open source hypervisors; Freet, Agrawal,
Walker and Badr [5] detail out the general characteristics that
give advantages of some over others. For example, their study
includes adoption reviews on Eucalyptus, OpenStack,
CloudStack, OpenNebula, Nimbus and Proxmox and
presents a conclusion that OpenStack and CloudStack have
over 30 times more messages in discussion forums that some
of their other competitors (meaning they are more top of mind
in the development community). They proceed to review the
architecture fit within three commercial offerings (Xen,
KVM, Virtual Box and ESX/VMware) in relation to the
requirements for data center virtualization and infrastructure
provision. In that study, various types of workloads are
simulated through each candidate hypervisor and the
throughput for each is reported. We have adopted a similar
throughput reporting in our methodology.
Vardhan Reddy and Rajamani [15] further study the
incremental overhead added by 4 different hypervisors. Their
work includes measurement of the residual CPU, memory,
I/O (read/write) and network (send/receive) with focused
workloads for Citrix XenServer, VMWare ESXi, Linux
(Ubuntu) KVM and Microsoft Hyper-V. They conclude that
the Hyper-V overall performance is very close to the winning
VMWare. Their results are useful as another data point for
our work (the work was done on a slightly older version than
ours). In our opinion, the Microsoft architecture has
continued to evolve in areas such as swap-file performance
and such stack would perform at least as well as their findings
indicate in similar tests. Their calculations on a 32GB cloud
indicate that there is a 30% overhead on RAM at that level.
Our experiments begin at 8GB memory allocation and they
increment by 2 GB in successive trials until the system
performance can be linearly approximated based on the
increments.
In yet a further diagnostic approach, Ye et. al. [16] propose
a very innovative method and system for measuring usage of
resources along the stack. They segment their findings into
impacts on hardware (indicating cache optimization should
be attempted), hypervisor (the overhead from the hypervisor
itself) and finally from the virtual machines themselves (the
workload profile). The Virt-B system reports the results from
these layers as various workloads are being processed. This
work not only quantifies the impacts on performance but
further diagnoses the parts of the stack that might have
significant bearing on the issue.
B. Virtualization Overhead Optimization
Virtualization of a platform’s resources can result in
significant incremental requirements compared to bare metal
8 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-20-320.jpg)
![architectures. There are however, a set of tools and
techniques that can help optimize those results in a virtualized
environment. Oi and Nakajima [8] explored the effects of
performance loads on the Xen. They determined that the
performance of Xen could be enhanced in a virtualized
environment by adjusting cache sizes in some applications in
addition to incremental memory. Virtualized-Xen and bare-
metal Linux were compared for throughput performance in
different cache and memory optimization techniques. In most
circumstances, it is a combination of both that will drive
throughput gains in a virtual environment. In their work, they
conclude that by varying configuration elements, a more
effective use of resources can be achieved. The
benchmarking system used was SPECjbb2001 and the effects
of Network Interface Cards (NIC) were isolated so the
workload could be measured in memory usage and
throughput. Our team has adopted the SPEC performance
suite as a workload simulator to determine the effects on
memory allocations (another attribute) rather than in network
throughput.
Another relevant study is Jabry, Liu, Zhu and
Panneerselvam [1]; hypervisor overhead impact is studied on
the following resources: disk i/o, CPU, memory and VMM
(hypervisor memory usage). The study predicts the usage of
resources by the hypervisor in taking total resource usage and
subtracting individual component loads and until only a
residual is left (presumably the hypervisor load). Those tests
were conducted with VMware, Virtual Box (Oracle
Corporation) and Windows Virtual PC (Microsoft). Their
work benchmarked a standard load in each hypervisor
environment and used IOzone to quantify load on disk i/o,
RAMSpeed to quantify the impact on memory and
UnixBench, to indicate the effect on CPU. Their work
concludes that the hypervisor is considerably higher on CPU
rather than the other components of the architecture. Each
suite of simulations focused on impacting a separate part of
the architecture and demonstrated how different workloads
impact the choice of hypervisor. It points to the Microsoft
stack being more balanced due to its integration with the
other components included in that specific Cloud architecture
(MS Windows). We selected the Microsoft stack in our
simulation so as to provide for greater integration between
the components and being able to evaluate the environment
as a “whole offering” from a single vendor. Further, tightness
of coupling between the units would allow for study of the
simulation as a whole without the need to study the effects of
separate vendor “noise”.
Chen, Patel, Shen and Zhou [2] studied virtualization
overhead across multiple VMs running under Xen in cloud
environments. They also found that the larger resource usage
was attributable to the CPU. They also propose a series of
equations that are remarkably accurate in predicting the
lateral scaling of workloads on all components based on the
observed results of the application under study. We provide a
graphical analysis of throughput under several memory
parameters (one of the parameters for optimization of CPU
performance).
C. Application Workload Research
Based on the research referenced, there is a significant
impact on utilization of CPU from the overhead generated by
the hypervisor. Further the impact is based on the type of
application that is operating in the virtualized environment.
NasiriGerdeh, Hosseini, RahimiZadeh and AnaLoui [7]
measured throughput degradation on Web applications using
the Faban suite (a web-based workload generator). They
simulated the behavior of heavy transactional Web
applications that tend to be very network intensive. Their
work also measured the effect on memory, disk i/o and CPU.
They concluded that a disproportionate difference exists in
CPU resources due to the translation of domain addresses.
This work further confirms that the principal resource
difference is the CPU utilization even when workloads may
be more i/o bound (the penalties associated where in finding
addresses; a CPU task, not access to the actual addresses in
the Web environment; an i/o task). We incorporate this
research by focusing on actual compute resource utilization
rather than network or disk access. The SPECjbb2015 suite
is focused on exhausting the compute resources rather than
the disk (i/o) or network resources.
San Wariya, Nair and Shiwani [11] focused their research
on benchmarking three hypervisors; Windows Hyper-V,
VMWare/ESXi and Citrix Xen in three cloud games; 3D
Mark 11, Unigine Heaven and Halo. The objectives of their
study are to identify which hypervisor was better from a
cloud gaming workload perspective. The three performed
differently in each category but were mostly lead by the
VMWare product. For our purposes however, the HALO
benchmark (number of frames per second) is probably the
most predictive of workloads that are CPU bound. In this
category, Hyper-V performed 7% ahead of VMWare and
57% ahead of Citrix Xen. This was another reason for
selection of Hyper-V as the hypervisor for our test suite.
D. The SPEC Benchmarking Suite
The SPECjbb2015 constitutes a workload benchmarking
simulation for a Supermarket Chain. The model can be
extended to include several supermarkets and several central
offices in a variety of virtual machine settings. The tool set
can be configured in a variety of business transaction settings
so that different business patterns can be simulated (e.g., web
sales versus physical store sales). The system is owned and
licensed by spec.org Error! Reference source not found.
which is a consortium of major IT companies that have
agreed on a set of principles to guide the performance
benchmarking process.
The system progressively injects transaction loads into the
environment until saturation is reached. A sample output of
these results is seen in Figure 3. In that graphic the system
begins to stress at around the 5,200 java Operations Per
Second (jOPS) with a range of 5K (median tolerance) to 50K
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 9
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-21-320.jpg)
![(max tolerance). The system reaches saturation (min
tolerance) at around 6,700 jOPS and 60K. We report our
results using the total transactions up to saturation. Figure 4
is a graphic representation of the architecture of the system.
Figure 3: Sample SPECjbb2015 Benchmark Output
www.spec.org
Figure 4: SPECjbb2015 Architecture
www.spec.org
The SPECjbb products have been in existence since the
late 1990s and are useful because of their industry
acceptance. For example, Karlsson, Moore, Hagersten and
Wood [6] used an earlier version (SPECjbb2001) and another
application benchmark (ECPerf) to differentiate effects of
cache misses between different types of applications.
III. EXPERIMENT DESIGN
As discussed above, experiments were designed where the
same application (SPECjbb2015) was installed on:
a) Bare metal with Windows 10
b) Bare metal with Windows Server 2012 R2
c) Virtual Machine: Windows Server 2012 R2 NOS /
Hyper-V/ Windows 10 Guest OS
The simulation was run for a typical store sales only
company with 90% store sales and 10% online sales. This is
typical of smaller stores that have not adapted to the online
grocery demands of consumers and are experimenting with
their own private clouds.
B. Application Architecture Patterns
The application patterns were analyzed by deriving use
cases and preparing activity diagrams from the code for the
application workload being simulated.
i. Use Case Analysis
Use cases are a functional decomposition tool that
illustrate the process interactions between actors in
applications [13]. The processes that we have selected in the
SPECjbb2015 suite are fairly standard and follow similar
patterns. The use case diagram for the store sales architecture
is similar to this one (www.UML-diagrams.org), the
“adornments” in the graphic describe the usage of artifacts.
Figure 5: Store Use Case Diagram
The inventory on-hand function at the physical store is
susceptible to over-booked demand and out-of-stock
conditions (where demand for an article exceeds supply). If
the system detects an out-of-stock condition, it will proceed
to cancel and back out the transaction. This process is
memory intensive since it has to place the order items back
into inventory and invalidate the order itself (see the error
exception in the UML activity diagram below).
ii. UML Activity Diagrams
UML activity diagrams are a useful tool for analyzing the
flow of logic through processes [13]. The following diagram
was created from the code in the application.
10 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-22-320.jpg)
![Figure 6: Store UML Activity Diagram
C. Bare Metal Implementations
Two bare metal SPECjbb2015 implementations were
used in the experiments; one on Windows 10 and one on
Windows Server 2012 R2. The allocations were 8GB, 10GB
and 12GB of memory (total of 6 bare metal simulations).
D. Virtualization Hypervisor Architecture
The third environment used MS Windows 10, Server 2012
R2 and Hyper-V [4]. The SPECjbb2015 software was
compiled inside the virtual machine (VM). A full physical
CPU, Network Interface Card (NIC) and all storage available
was allocated to the VM. RAM of 8GB, 10GB and 12GB was
allocated to the Virtual Machine. A diagram for this
architecture is shown below, (ours has one VM).
Figure 7: Virtualized Environment Architecture
(www.microsoft.com)
E. Infrastructure (Machine) Specifications
The infrastructure environment that the experiments were
executed on had the following specifications:
• Hewlett-Packard Envy 15t
• Intel i6700 quad-core processor
• 16G RAM
• 1TB Hybrid SSD
• 4GB NVIDIA GTX 950M chip
IV. EXPERIMENT RESULTS
Figure 8 reports the total transaction throughput achieved
under the different memory allocations in each environment.
Figure 9 reports the percentage increase, using the 8GB results
as the base for each environment. This is done to stress the
incremental impact of each increase on the original base
measurements.
Figure 8: Throughput in the Experimental Systems
The above indicates there were marginal increases in
throughput on bare metal versus significant increases with
increase in memory on the virtual machine.
Figure 9: Percentage Gain in Throughput for the
Experimental Systems
The above emphasizes the gains throughput when
memory is increased in the virtual machine. Throughput
keeps increasing at significant rates (although begins to curb
with the second increase in memory to 12GB).
V. DISCUSSION & FUTURE PLANS
The objectives of this study are to isolate the impact of
additional memory allocation on a static workload. The
hypothesis that additional memory increases throughput in
virtualized environments. Part of this benefit is slowed as the
allocation progresses.
A. Results Discussion
As systems are virtualized, they consume greater
resources due to “virtualization overhead”; they require
translation of the logical to the physical and back to the
start
Txi Sends SM an
InStorePurchaseRequest
Select a Random
Customer
Retrieve Customers
Previous Purchase History
Reserve Specific
Quantity Of Each
Product
Calculate Total price
Add the Available
Discounts and
Coupouns
Customer Basket Validation
Max Products
Available
Many Products to
be Replenished
Throw an Exception
Proceed to Check
out
Generate Reciept
Check Customer's Credit
Customer has
enough Credit
Customer doesn't
have enough Credit
STOP
Transaction Fails
Move the Purchased
Items from Store
Inventory
Debit the cost of Each
Item from Customer's
Account
Send Suppliers a
Request, If any item runs
out from store
Send Receipt back
to HQ
Stop
Instore
Activity
10GB 12GB
Win 10 2% 2%
Win 2012 20% 7%
Win Hyp-V 95% 79%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Percent Throughput Improvement, 8GB Constant Base
10GB 12GB
Win 10 2% 2%
Win 2012 20% 7%
Win Hyp-V 195% 37%
0%
50%
100%
150%
200%
250%
Percent Throughput Improvement, 8GB Constant Base
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 11
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-23-320.jpg)
![logical. Our study has illustrated how using a process of
simulation, a small company may avoid the costly and risky
process of making the decision to virtualize in a private cloud
without knowing how much incremental resource will be
needed.
We used Windows 10 and Server 2012 R2 to test on bare
metal. It is important to understand that these are two
fundamentally different Operating Systems. Windows 10 is
focused on managing the desktop and execution of localized
workloads. It provides rich functionality in areas such as
graphics and gaming which are not simulated by the
SPECjbb2015 suite but are nevertheless instantiated in it’s
services. The Server 2012 R2 is a Distributed OS whose focus
is to manage standard workloads associated with raw
compute and storage power. Under these circumstances, the
Server 2012 R2 performs with better results in traditional
business process simulations like SPECjbb2015.
The Hyper-V extension of the Windows Server 2012 R2
is a tool for managing the life cycle of virtual machines. It is
an extension of the Network OS that communicates decisions
to the virtualization layer for translation of operating
parameters back and forth. This additional load constitutes
overhead (more resources). Some of these resources are
“fixed”; they are there by virtue of instantiation and some are
“variable”; by usage of the workspace through applications.
As memory is increased, there exists more workspace for
applications and the overall impact of the hypervisor usage of
memory is reduced. The first incremental memory allocation
(from 8GB to 10GB) has a higher yield because a greater
percentage of that “boost” goes to the application. The
successive increment (from 10GB to 12GB) is still significant
but not as high. According to hypervisor vendors and Reddy
and Rajamani [15] this reduction will continue until the bare
metal and virtualized environment will start to resemble each
other in throughput given volume/memory mix.
The simulation process allows a company to make plans
of how to deploy in the future. Using a standard simulation
may lead to answering some key questions as:
a) Should we virtualize or keep on bare metal?
b) When should we revisit our decision?
A company could continue the simulation by contracting
additional capacity with one of the major Cloud vendors and
determine where the VM/bare metal results ultimately blur.
B. Future Plans
The team is busy executing additional work in running
additional simulations that can implement optimization
techniques. The ultimate objective is to have a “cookbook” of
simulation/optimization techniques that can be used in private
or hybrid cloud evaluators.
REFERENCES
[1] H. Al Jabry; L. Liu, Y. Zhu, J. Panneerselvam: “A
Critical Evaluation of the Performance of Virtualization
Technologies”. 9th
International Conference on
Communications and Networking in China (2014). p.
606-611.
[2] L. Chen; S. Patel; H. Shen; Z. Zhou: “Profiling and
Understaning Virtualization Overhead in Cloud”. 2015
44th
International Conference on Parallel Processing. p.
31-40.
[3] T. Erl; Z. Mahmood; R. Puttini: “Cloud Computing
Concepts, Technology & Architecture”. c. 2013 Arcitura
Education, Inc./Pearson Education, Upper Saddle River,
NJ. USA.
[4] A. Finn; M. Luescher; P. Lownds; D. Flynn: “Windows
Server 2012 Hyper-V; Installation and Configuration
Guide”. c. 2013 Wiley and Sons, Indianapolis, IN. USA.
[5] D. Freet; R. Agrawal; J. Walker; Y. Badr: “Open source
cloud management platforms and hypervisor
technologies: A review and comparison”. SoutheastCon
2016. p. 1-8.
[6] M. Karlsson; K.E. Moore; E. Hagersten; D.A. Wood:
“Memory system Behavior of Java-Based Middleware”.
The Ninth International Symposium on High-
Performance Computer Architecture, 2003. HPCA-9
2003 p. 217-228.
[7] R. NasiriGerdeh; N. Hosseini; K. RahimiZadeh; M.
AnaLoui: “Performance Analysis of Web Application in
Xen-based Virtualized Environment”. 2015 5th
International Conference on Computer and Knowledge
Engineering (ICCKE). p.258-261.
[8] H. Oi; F. Nakajima: “Performance Analysis of Large
Receive Offload in a Xen Virtualized System”. 2009
International Conference on Computer Engineering and
Technology. p. 475-480
[9] J. Rhoton; J. De Clercq; F. Novak: “OpenStack Cloud
Computing Architecture Guide 2014 Edition”. c. 2014
Recursive Press, USA & UK.
[10] A. Salam; Z. Gilani; S. Ul Haq: “Deploying and
Managing a Cloud Infrastructure”. c. 2014 Sybex, a
Wiley Brand, Indianapolis, IN. USA
[11] A. SanWariya; R. Nair; S. Shiwani: “Analyzing
Processing Overhead of Type-0 Hypervisor for Cloud
Gaming”. 2016 International Conference on Advances in
Computing, Communication, & Automation (ICACCA)
(Spring). p. 1-5.
[12] W. Stallings: “Operating Systems: Internals and Design
Principles 7th
ed”. c. 2012 Pearson/Prentice Hall, Upper
Saddle River, NJ. USA
[13] C. Larman; “Applying UML and Patterns: An
Introduction to Object-Oriented Analysis and Design
and Iterative Development (3rd Edition)” c. Pearson
Education 2005, Upper River, NJ.
[14] www.spec.org
[15] P.V. Vardhan Reddy; L. Rajamani: “Virtualization
Overhead Findings of Four Hypervisors in the
CloudStack with SIGAR”. 2014 World Congress on
Information and Communication Technologies (WICT
2014) p. 140-145.
[16] K. Ye; Z. Wu; B. Zhou; X. Jiang; C. Wang; A. Zomaya:
“Virt-B: Toward Performance Benchmarking of Virtual
Machine Systems”. IEEE Internet Computing, V. 18,
Issue 3 (2014). p. 64-72
12 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-24-320.jpg)
![Autonomously Battery Charging Tires For EVs Using
Piezoelectric Phenomenon
Muhammad Kamran1
, Dr. Raziq Yaqub2
, Dr. Azzam ul Asar1
1
CECOS University, Peshawar, Pakistan, 2
University of Tennessee, USA
Abstract:-This paper illustrates the use of piezoelectric
material to generate electricity in the Electric Vehicle (EV)
tire. According to the proposed mechanism, the vehicle tires
are embedded with layers of the Piezo-electric material
along the periphery. Thus, when the EV is in motion,
electricity can be generated due to mechanical stress in that
part of Piezo-electric material that is in contact with the
road surface. The results show that, with peripheral
arrangement of Piezo-material inside the automobile tire,
we can generate electricity that can be stored in a battery to
run the EV for some extra miles of the total miles the EV is
capable of on a single charge. The use of Polyvinylidene
fluoride (PVDF); a polymer based piezoelectric material is
considered due to its robust and favorable properties.
Keywords— Piezoelectricity, Polyvinylidene fluoride
(PVDF), Electrical power, Mechanical stress-to-electricity
conversion, Automobile tire.
1 Introduction
Due to the rising demand for generating energy in the
most efficient way; smart, intelligent and adaptive materials
are being used and one such smart substance is the
piezoelectric material. Piezoelectric substances produce
electric charge when mechanical stress is applied on its
surface.
Piezoelectric materials are composed of various
materials namely crystals, ceramics, polymers etc. Polymer-
based piezoelectric materials have served as the most
efficient material compared to ceramics and crystals for
applications where elasticity is preferred. The most
commonly used polymer based piezoelectric material is
Polyvinylidene Fluoride (PVDF). PVDF is a transparent,
semi-crystalline, thermoplastic fluoroplastic. We have
employed PVDF as the piezo electric material in our work
based on the merits of PVDF as listed below [13]:
(i) Piezoelectricity obtained from PVDF is several times
greater than that obtained from quartz or ceramics.
(ii) PVDF materials are insoluble in water, resistant to
solvents, acids, bases, heat, and generate low smoke in case
of any fire accidents.
(iii) Has low weight and low thermal conductivity.
(iv) Highly resistant to chemical corrosion and heat
variations, thus withstands exposure to harsh chemical and
thermal conditions.
(v) Very good mechanical strength and toughness and has
high abrasion resistance.
(vi ) Low permeability to most gases and liquids.
(vii) Unaffected by long-term exposure to ultraviolet
radiation.
(viii) Less expensive compared to its counterpart.
These features make them most suited to be employed in
EV tires. However, proof of concept needs to be done, that
requires collaboration with tires manufacturer.
The rest of the paper is divided into the following
sections. Section-2 describes the composition and structure
of the tire, section-3 explians our proposal on embedding
piezo-electric material in tire and harvesting energy from it.
Section-4 calculates cost efficiency of the proposed
mechanism, Section-5 suggest future work, Section-6
concludes the paper, Section-7 lists some of the key
references, and finally Section-8 titled as Annex, presents
the detailed mathematical analysis.
Due to the rising demand for generating energy in the
most efficient way; smart, intelligent and adaptive materials
are being used and one such smart substance is the
piezoelectric material. Piezoelectric substances produce
electric charge when mechanical stress is applied on its
surface.
Piezoelectric materials are composed of various
materials namely crystals, ceramics, polymers etc. Polymer-
based piezoelectric materials have served as the most
efficient material compared to ceramics and crystals for
applications where elasticity is preferred. The most
commonly used polymer based piezoelectric material is
Polyvinylidene Fluoride (PVDF). PVDF is a transparent,
semi-crystalline, thermoplastic fluoroplastic. We have
employed PVDF as the piezo electric material in our work
based on the merits of PVDF as listed below [13]
2 Composition and Structure of the Tire
The most basic component in the tire is “rubber” which
may be ‘synthetic rubber’ or ‘natural rubber’. Other
components that are present in the tire are fabric wire,
polymers, fillers, softeners, anti-degradents and curatives.
As polymers are the backbone of rubber compounds, it is
more appropriate to embed polymer piezoelectric material
within the structure of the tire as done in [1].
Since the objective of using Piezoelectric material in [1]
is sensing, it simply employs pallets of PVDF materials.
However, we embed PVDF material as a circular ring along
the entire periphery of the tire to maximize electricity
generation.
There are 3 main categories of tires such as Diagonal
(bias) tire, belted bias tire and Radial tire. Radial tires are
most commonly used in the automobile industry; therefore,
this paper considers radial tire for mathematical analysis.
However, it does not preclude other types.
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 13
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-25-320.jpg)
![Fig. 2 Piezoelectric Material as a Circular Ring in a radial tire
3 Proposal on Embedding Piezoelectric
Material in Tires to Harvest Energy
Figure 1 shows the overall concept of piezoelectric
generation phenomenon in context to the proposed tire
scenario.
Fig. 1. Overall Concept of Piezoelectric Generation in Proposed Tire.
Figure 2 shows the cross-sectional area of the original
radial tire, where we proposed to have one layer of
piezoelectric material, along the periphery of the tire, below
the rubber layer or any suitable place, the tire manufacturers
deem suitable. Having the PVDF material as a circular ring
along the entire periphery of the tire is considered to be
more efficient compared to having pallets of PVDF material
embed within the tire. It is because when the tires will
rotate, most portions of the PVDF will be coming in contact
with the road, and thus will ensure constant generation of
electricity.
Electricity is generated in piezoelectric materials due to
mechanical stress in that part of piezoelectric material that
is in contact with the road surface. Piezoelectricity is the
direct result of the piezoelectric effect. The electricity so
produced is fed to the car battery.
Having the PVDF material as a circular ring along the
entire periphery of the tire is more efficient compared to
having pallets of PVDF material embed within the tire.
When the tires will rotate at a high speed, most portions of
the PVDF will be coming in contact with the road, and thus
will ensure constant generation of electricity as shown in
figure 1.
In this section, pressure exerted by the road on the
automobile tire is modeled and the amount of energy
harvested in this process is calculated. Since the section
involves mathematical variables, therefore for the
convenience of the readers, the terminology, abbreviations
and units are provided in the form of a table below:
TABLE I. NOMENCLATURE, SYMBOLS AND UNITS
Pa
ra-
meter
Parameter definition Unit
A Area meter2
c Circumference centimeter
C Charge Coulomb
d Piezoelectric strain
coefficient or
D Diameter centimeter
F Force Newton
g Gravity meter/sec2
I Current Ampere
k Distance centimeter
m Mass of the car kilogram
Pp Charge surface density Coulomb/m2
p Power Watts
P Pressure Newton/m2
t1 Time seconds
T Pressure exerted on the
PVDF material
Newton/met
er2
v Velocity/speed Miles/hour
V Voltage volts
W Width Centimeter
g Appropriate piezoelectric
coefficient for the axis of
applied stress or strain
or
t Thickness of a ring µm
Modeling the System
The experiments performed by the Curie brothers
demonstrated that the Charge Surface Density is
proportional to the pressure exerted, and is given by [2]
Pressure exerted by a car on the road can be given by,
Tire With
Piezoelectric
Material
Road Surface
Piezoelectric
Phenomenon
Tire With
Piezoelectric
Material
Road Surface
Piezoelectric
Phenomenon
14 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-26-320.jpg)
![Where F is the force exerted on the tire and is equal to
the weight (F = mg, where m is the mass of the car and g is
the force of gravity and is 9.8 m/s2
). And A is the tire
surface area in contact with the road.
The tire surface area A in contact with the road can be
calculated as
A = π x D x W x 0.1
Where 0.1 is due to the fact that 10% of the tire area is
in contact with the surface of the road [3].
Output voltage for the given stress or strain is given by
V0 = g3n x Xn x t
Where n = 1, 2 or 3 and Xn = T [7].
As we are considering n=1 i.e. (in the piezoelectric
material the electrical axis is always fixed as it is three in
this case and the mechanical axis is either one, two or
three), the value of g31 is specified in table 2. Moreover we
are considering force on tire due to weight along Y direction
as shown in figure. The component of weight acting in this
way is constant and 20% of total weight [14] [15]. Now if
the force is assumed to be acting axially then the area
should also be taken in the specified direction. So the V0
will be modified as [7].
V0 = g31 × (Force/width × thickness) × thickness
= g31 × (Force/width)
Fig 3. [15]
Fig. 4 [15]
Fig. 5. [15]
Result Calculation
If the mass of the car is 1500Kg (Because electric
vehicles have more weight due to their battery) then the
calculated force is 14700Newtons (3675N for each tire).
When the average diameter of the PVDF ring is considered
to be D = 0.5588 m (22 inch), width of a ring is 0.1651 m
(6.5 inch) and thickness of the ring is 110 micrometer [7].
Therefore, the Area (A) for the PVDF ring is 0.289 m2
and
the surface area when 10% of the tire is in contact with the
road surface is 28.9 × 10-3
m2
.
From this value of force (F) and area (A) the pressure
(T) exerted on the PVDF material is 127162.63 N/m2.
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 15
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-27-320.jpg)
![TABLE II. TYPICAL PROPERTIES OF PVDF [7]
Parameter Sy
mbol
Valu
e
Unit
Thickness t 55 µm
(micron, 10-6
)
Piezoelectric
Strain Constant
d31 23 10-12
or
Piezoelectric
Stress Constant
g31 216 10-3
or
According to above PVDF properties and equation the
charge surface density is 2.92 × 10-6
C/m2
.
We know that the,
Charge = Current x Time,
and
Power = Current x Voltage
Assumption has been made in order to make
calculations easier, that the amount of time taken to
generate the electricity is one second, allowing the charge to
be equal to the current [3].
From all of the above equations and assumptions the
power generated is 2.81mW for 1 tire. We consider only
10% of the tire touches the road surface, so in one rotation
10 times electricity is generated. In one rotation 28.1mW
power is generated.
The US environment protection agency official range is
117 km (73 mile) with an energy consumption of 765
kilojoules per kilometer or 34 KWh/100mile [6] [12].
According to this energy consumption our designed car can
run extra 39 km because the power generated by four tires
of the car with any specified speed.
4 Cost Efficiency
Cost of the tire after incorporating the PVDF ring inside
the tire depends mainly on two factors, cost of the PVDF
and the cost for implementing the PVDF ring inside the tire.
The PVDF ring cost is from $50 - $100 per Cubic Meter,
which depends on length, width and thickness of the ring
[4].
If we consider the cost of the PVDF to be used in four
tires to be $50 including embedding process of PVDF
material inside the tire. The average life of all-season radial
tire advertised by the manufacturer is 50000 miles. Using
proposed technology, the EV can bring a cost saving worth
17500 additional miles. If the cost of electrical energy is
$0.04/mile, a saving of $525 can be achieved.
5 Future Work
We plan to do the following work in future: Followings
tasks would be carried throughout the length of the project.
(i) Produce simulations considering the tire industry
standards such as tread, the body with sidewalls, and
the beads (the rubber-covered, metal-wire that hold the
tire on the wheel)
(ii) Use Autodesk (It is already licensed to UT) sofware to
simulate the performance of design parameters such as
distribution of PVDF around the periphery of the tire.
(iii) Simulate the effects of different typs/concentrations of
PVDF compounds and different types of distribution of
PVDF around the periphery of the tire. And also using
different concentrations of PVDF in different parts of
the tire.
(iv) Simulate the effects of different typs/concentrations of
rubber compounds and different types of distribution of
rubber around the periphery of the tire. And also using
different concentrations/ratios of rubber and PVDF in
different parts of the tire.
(v) Analyze the effects of different stresses on the proosed
tire design (emulating different weights, road
roughness, etc.), and discover design limitations.
(vi) Analyze the effect of different temperatures (eulating
different hot/cold weathers). Also simulate the effects
of different sizes the tires come in.
It has to be ensured that simulations as well as timelines
for simulations meet the expectations, through validation,
and comparison with the specifications of standard tires
(non PVDF tires).
6 Conclusion
Our work demonstrates a method of generating
electricity using the PVDF material. Mathematical analysis
proves that the EV can run extra 37 miles on a single charge
with a speed of the car is 60mph. Since the cost of PVDF
and its implementation is not so expensive, a saving of
about $500 is expected over the life of the tire. Overall, the
proposed method is an excellent choice to generate power
when the car is on move.
Acknowledgment
The 1st
and 3rd authors would like to acknowledge the
technical support of Dr. Raziq Yaqub for his valuable
contribution extended during the course of this project and
allowing to improve the mathematical model.
References
[1] Jingang Yi, "A Piezo-Sensor-Based 'Smart Tire' System
for Mobile Robots and Vehicles",March 2007.
[2] Arnau, Antonio. “Fundamentals on Piezoelectricity.”
Piezoelectric Tranducers and Applications. New York,
2008. Print, pp.4.
[3] http://cosmos.ucdavis.edu/archives/2011/cluster2/Yau_
Derek.pdf.
[4] http://www.alibaba.com/product-
gs/322181211/PVDF_Intalox_Saddle_Ring.html
[5] http://www.carfolio.com/specifications/models/car/car=
107844&GM.
16 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 |
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-28-320.jpg)
![[6] http://www.autoblog.com/2009/08/01/2010-nissan-leaf-
electric-car-in-person-in-depth-and-u-s-b/
[7] Measurement Specialities Inc. April 1999 “Piezo Film
Sensor Technical Manual”, P/N 1005663-1, Rev. B, pp.
3-4,28
[8] West Coast Green Highway. Electric Highways
Project, 2010. Retrieved March 09, 2012, from
http://west
coastgreenhighway.com/electrichighways.htm
[9] Kris De Decker, May 3, 2010. “The status quo of
electric cars: better batteries, same range.” Low-tech
magazine. Retrieved March 13, 2012 from
http://www.energybulletin.net/node/52736.
[10] http://www.tirerack.com/tires/tiretech/techpage.jsp?tec
hid=46
[11] http://www.tirerack.com/tires/tiretech/techpage.jsp?tec
hid=7.
[12] http://teacher.pas.rochester.edu/phy121/lecturenotes/Ch
apter06/Chapter6.html.
[13] http://www.openmusiclabs.com/wp/wp-
content/uploads/2011/11/piezo.pdf
[14] http://en.wikipedia.org/wiki/Nissan_Leaf
[15] http://en.wikipedia.org/wiki/Polyvinylidene_fluoride
[16] http://www.mate.tue.nl/mate/pdfs/8351.pdf
[17] http://road-transport-
technology.org/Proceedings/2%20-
%20ISHVWD/Vol%201/TRUCK%20TIRE%20TYPE
S%20AND%20ROAD%20CONTACT%20PRESSUR
ES%20-%20Yap%20.pdf
[18] http://www.tzlee.com/blog/?m=201103
Annex: Detailed Mathematical Analysis
Calculating Patch Area of PVDF Ring
Contact patch (also called footprint) is the area in which
the tire is in contact with the road surface). Different
vehicles have different contact patch depending on tire’s
diameter and width. Tires diameter ranging from 8 to 26
which are given in detail in [8] [9]. For the sake of analysis
we considered a tire with the diameter of 22 inches and
width of 6.5. (I.e. the tire size of 185/55R15 commonly used
for Passenger Electric Vehicles (EV)). We consider
incorporating PVDF ring inside the whole width of tire, so
that the tire continue to adhere with its original texture
without scarifying its original purpose or violating its
specifications in terms of road resistance, air pressure, etc.
D = 0.5588 meters
W = 0.1651 meters
Therefore, Area (A) is given by,
ATotal = ∏ × D × W
= ∏ × 0.5588 × 0.1651
= 0.289 m2
We assume that only 10% of the area is in contact with
the road surface. Therefore, the Area (A) is given by,
A = 28.9 × 10-3 m2
To determine the Pressure
If we consider the mass of the car to be 1500 kg, force
can be calculated as follows:
Force = m × g
= 1500 × 9.8 = 14700 N
And
Force on one tire will be = 14700/4 =3675N
Therefore Pressure = 14700 / 28.9 × 10-3 = 127162.63
N / m2
To determine the charge surface density
The charge surface density is given by,
Where d is the piezoelectric strain coefficient and from
table 2, its value is given to be 23 × 10-12. Therefore,
= 23 × 10-12 × 127162.63
= 2.92 × 10-6 C/m2
To determine the output voltage
In the previous section, we have briefly discussed the
equations for
Output voltage which is given by,
V0 = g3n x Xn x t
In our work we consider n = 1 and the value of g31 and t
is specified in table 2. Xn = (Force/width × thickness)
Therefore, output voltage is,
V0 = g31 × (Force/width × thickness) × thickness
= g31 × (Force/width)
= 216 × 10-3
× (0.2×3675/0.1651)
= 961.6 V
And 0.2 is the component of force that acts axially
shown in figure-2c, as we are considering g31 mode so the
force should also be considered along the specified
direction.
To determine the Total Power
Total Power = Charge Surface Density x Output voltage
= 2.92 × 10-6 × 961.6 = 2.81 mW
We consider only 10% of the tire touches the road
surface, so in one rotation 10 times electricity is generated.
In one rotation of one tire 28.1mW power is produced.
Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 17
ISBN: 1-60132-465-0, CSREA Press ©](https://image.slidesharecdn.com/54033-250512223838-456939e6/85/Modeling-Simulation-and-Visualization-Methods-1st-Edition-Hamid-R-Arabnia-29-320.jpg)
Modeling Simulation and Visualization Methods 1st Edition Hamid R. Arabnia
- 1. Modeling Simulation and Visualization Methods 1st Edition Hamid R. Arabnia download https://ebookname.com/product/modeling-simulation-and- visualization-methods-1st-edition-hamid-r-arabnia/ Get Instant Ebook Downloads – Browse at https://ebookname.com
- 2. Instant digital products (PDF, ePub, MOBI) available Download now and explore formats that suit you... Information and Knowledge Engineering 1st Edition Hamid R. Arabnia https://ebookname.com/product/information-and-knowledge- engineering-1st-edition-hamid-r-arabnia-2/ Information and Knowledge Engineering 1st Edition Hamid R. Arabnia https://ebookname.com/product/information-and-knowledge- engineering-1st-edition-hamid-r-arabnia/ Software Engineering Research and Practice 1st Edition Hamid R. Arabnia https://ebookname.com/product/software-engineering-research-and- practice-1st-edition-hamid-r-arabnia/ Private 2 Invitation Only Kate Brian https://ebookname.com/product/private-2-invitation-only-kate- brian/
- 3. Bar Style Hotels and Members Clubs 1st Edition Howard Watson https://ebookname.com/product/bar-style-hotels-and-members- clubs-1st-edition-howard-watson/ Common Pools of Genetic Resources Equity and Innovation in International Biodiversity Law 1st Edition Evanson Chege Kamau https://ebookname.com/product/common-pools-of-genetic-resources- equity-and-innovation-in-international-biodiversity-law-1st- edition-evanson-chege-kamau/ The Conch Bearer 1st Aladdin Paperback ed Edition Chitra Banerjee Divakaruni https://ebookname.com/product/the-conch-bearer-1st-aladdin- paperback-ed-edition-chitra-banerjee-divakaruni/ The Working Class Majority America s Best Kept Secret 2nd Edition Michael Zweig https://ebookname.com/product/the-working-class-majority-america- s-best-kept-secret-2nd-edition-michael-zweig/ Private Lives Public Deaths Antigone and the Invention of Individuality 1st Edition Jonathan Strauss https://ebookname.com/product/private-lives-public-deaths- antigone-and-the-invention-of-individuality-1st-edition-jonathan- strauss/
- 4. Fundamentals Of Polymer Engineering 2nd ed, rev. and expanded Edition Anil Kumar https://ebookname.com/product/fundamentals-of-polymer- engineering-2nd-ed-rev-and-expanded-edition-anil-kumar/
- 5. Editors Associate Editors © CSREA Press Ashu M. G. Solo, Jane You PROCEEDINGS OF THE 2017 INTERNATIONAL CONFERENCE ON MODELING, SIMULATION & VISUALIZATION METHODS Hamid R. Arabnia Leonidas Deligiannidis, Fernando G. Tinetti CSCE’17 July 17-20, 2017 Las Vegas Nevada, USA americancse.org
- 6. Copyright and Reprint Permission Copying without a fee is permitted provided that the copies are not made or distributed for direct commercial advantage, and credit to source is given. Abstracting is permitted with credit to the source. Please contact the publisher for other copying, reprint, or republication permission. Copyright © 2017 CSREA Press ISBN: 1-60132-465-0 Printed in the United States of America This volume contains papers presented at The 2017 International Conference on Modeling, Simulation & Visualization Methods (MSV'17). Their inclusion in this publication does not necessarily constitute endorsements by editors or by the publisher.
- 7. Foreword It gives us great pleasure to introduce this collection of papers to be presented at the 2017 International Conference on Modeling, Simulation and Visualization Methods (MSV’17), July 17-20, 2017, at Monte Carlo Resort, Las Vegas, USA. An important mission of the World Congress in Computer Science, Computer Engineering, and Applied Computing, CSCE (a federated congress to which this conference is affiliated with) includes "Providing a unique platform for a diverse community of constituents composed of scholars, researchers, developers, educators, and practitioners. The Congress makes concerted effort to reach out to participants affiliated with diverse entities (such as: universities, institutions, corporations, government agencies, and research centers/labs) from all over the world. The congress also attempts to connect participants from institutions that have teaching as their main mission with those who are affiliated with institutions that have research as their main mission. The congress uses a quota system to achieve its institution and geography diversity objectives." By any definition of diversity, this congress is among the most diverse scientific meeting in USA. We are proud to report that this federated congress has authors and participants from 64 different nations representing variety of personal and scientific experiences that arise from differences in culture and values. As can be seen (see below), the program committee of this conference as well as the program committee of all other tracks of the federated congress are as diverse as its authors and participants. The program committee would like to thank all those who submitted papers for consideration. About 65% of the submissions were from outside the United States. Each submitted paper was peer-reviewed by two experts in the field for originality, significance, clarity, impact, and soundness. In cases of contradictory recommendations, a member of the conference program committee was charged to make the final decision; often, this involved seeking help from additional referees. In addition, papers whose authors included a member of the conference program committee were evaluated using the double-blinded review process. One exception to the above evaluation process was for papers that were submitted directly to chairs/organizers of pre-approved sessions/workshops; in these cases, the chairs/organizers were responsible for the evaluation of such submissions. The overall paper acceptance rate for regular papers was 25%; 19% of the remaining papers were accepted as poster papers (at the time of this writing, we had not yet received the acceptance rate for a couple of individual tracks.) We are very grateful to the many colleagues who offered their services in organizing the conference. In particular, we would like to thank the members of Program Committee of MSV’17, members of the congress Steering Committee, and members of the committees of federated congress tracks that have topics within the scope of MSV. Many individuals listed below, will be requested after the conference to provide their expertise and services for selecting papers for publication (extended versions) in journal special issues as well as for publication in a set of research books (to be prepared for publishers including: Springer, Elsevier, BMC journals, and others). • Prof. Nizar Al-Holou (Congress Steering Committee); Professor and Chair, Electrical and Computer Engineering Department; Vice Chair, IEEE/SEM-Computer Chapter; University of Detroit Mercy, Detroit, Michigan, USA • Prof. Hamid R. Arabnia (Congress Steering Committee); Graduate Program Director (PhD, MS, MAMS); The University of Georgia, USA; Editor-in-Chief, Journal of Supercomputing (Springer);Fellow, Center of Excellence in Terrorism, Resilience, Intelligence & Organized Crime Research (CENTRIC). • Prof. Dr. Juan-Vicente Capella-Hernandez; Universitat Politecnica de Valencia (UPV), Department of Computer Engineering (DISCA), Valencia, Spain • Prof. Juan Jose Martinez Castillo; Director, The Acantelys Alan Turing Nikola Tesla Research Group and GIPEB, Universidad Nacional Abierta, Venezuela • Prof. Kevin Daimi (Congress Steering Committee); Director, Computer Science and Software Engineering Programs, Department of Mathematics, Computer Science and Software Engineering, University of Detroit Mercy, Detroit, Michigan, USA • Prof. Zhangisina Gulnur Davletzhanovna (IPCV); Vice-rector of the Science, Central-Asian University, Kazakhstan, Almaty, Republic of Kazakhstan; Vice President of International Academy of Informatization, Kazskhstan, Almaty, Republic of Kazakhstan
- 8. • Prof. Leonidas Deligiannidis (Congress Steering Committee); Department of Computer Information Systems, Wentworth Institute of Technology, Boston, Massachusetts, USA; Visiting Professor, MIT, USA • Prof. Mary Mehrnoosh Eshaghian-Wilner (Congress Steering Committee); Professor of Engineering Practice, University of Southern California, California, USA; Adjunct Professor, Electrical Engineering, University of California Los Angeles, Los Angeles (UCLA), California, USA • Prof. Byung-Gyu Kim (Congress Steering Committee); Multimedia Processing Communications Lab.(MPCL), Department of Computer Science and Engineering, College of Engineering, SunMoon University, South Korea • Prof. Dr. Guoming Lai; Computer Science and Technology, Sun Yat-Sen University, Guangzhou, P. R. China • Prof. Hyo Jong Lee (IPCV); Director, Center for Advanced Image and Information Technology, Division of Computer Science and Engineering, Chonbuk National University, South Korea • Dr. Muhammad Naufal Bin Mansor; Faculty of Engineering Technology, Department of Electrical, Universiti Malaysia Perlis (UniMAP), Perlis, Malaysia • Dr. Andrew Marsh (Congress Steering Committee); CEO, HoIP Telecom Ltd (Healthcare over Internet Protocol), UK; Secretary General of World Academy of BioMedical Sciences and Technologies (WABT) a UNESCO NGO, The United Nations • Prof. Aree Ali Mohammed; Head, Computer Science Department, University of Sulaimani, Kurdistan Region, Iraq • Dr. Ali Mostafaeipour; Industrial Engineering Department, Yazd University, Yazd, Iran • Prof. Dr., Eng. Robert Ehimen Okonigene (Congress Steering Committee); Department of Electrical & Electronics Engineering, Faculty of Engineering and Technology, Ambrose Alli University, Edo State, Nigeria • Prof. James J. (Jong Hyuk) Park (Congress Steering Committee); Department of Computer Science and Engineering (DCSE), SeoulTech, Korea; President, FTRA, EiC, HCIS Springer, JoC, IJITCC; Head of DCSE, SeoulTech, Korea • Prof. Dr. R. Ponalagusamy; Department of Mathematics, National Institute of Technology, India • Dr. Xuewei Qi; Research Faculty & PI, Center for Environmental Research and Technology, University of California, Riverside, California, USA • Dr. Akash Singh (Congress Steering Committee); IBM Corporation, Sacramento, California, USA; Chartered Scientist, Science Council, UK; Fellow, British Computer Society; Member, Senior IEEE, AACR, AAAS, and AAAI; IBM Corporation, USA • Ashu M. G. Solo (Publicity), Fellow of British Computer Society, Principal/R&D Engineer, Maverick Technologies America Inc. • Prof. Dr. Ir. Sim Kok Swee; Fellow, IEM; Senior Member, IEEE; Faculty of Engineering and Technology, Multimedia University, Melaka, Malaysia • Prof. Fernando G. Tinetti (Congress Steering Committee); School of CS, Universidad Nacional de La Plata, La Plata, Argentina; Co-editor, Journal of Computer Science and Technology (JCS&T). • Dr. Haoxiang Harry Wang (CSCE); Cornell University, Ithaca, New York, USA; Founder and Director, GoPerception Laboratory, New York, USA • Prof. Shiuh-Jeng Wang (Congress Steering Committee); Director of Information Cryptology and Construction Laboratory (ICCL) and Director of Chinese Cryptology and Information Security Association (CCISA); Department of Information Management, Central Police University, Taoyuan, Taiwan; Guest Ed., IEEE Journal on Selected Areas in Communications. • Prof. Layne T. Watson (Congress Steering Committee); Fellow of IEEE; Fellow of The National Institute of Aerospace; Professor of Computer Science, Mathematics, and Aerospace and Ocean Engineering, Virginia Polytechnic Institute & State University, Blacksburg, Virginia, USA • Prof. Jane You (Congress Steering Committee); Associate Head, Department of Computing, The Hong Kong Polytechnic University, Kowloon, Hong Kong We would like to extend our appreciation to the referees, the members of the program committees of individual sessions, tracks, and workshops; their names do not appear in this document; they are listed on the web sites of individual tracks. As Sponsors-at-large, partners, and/or organizers each of the followings (separated by semicolons) provided help for at least one track of the Congress: Computer Science Research, Education, and Applications Press (CSREA); US Chapter of World Academy of Science; American Council on Science & Education & Federated Research Council (http://www.americancse.org/); HoIP, Health Without Boundaries, Healthcare over Internet Protocol, UK (http://www.hoip.eu); HoIP Telecom, UK (http://www.hoip-telecom.co.uk); and WABT, Human Health Medicine, UNESCO NGOs, Paris, France
- 9. (http://www.thewabt.com/ ). In addition, a number of university faculty members and their staff (names appear on the cover of the set of proceedings), several publishers of computer science and computer engineering books and journals, chapters and/or task forces of computer science associations/organizations from 3 regions, and developers of high-performance machines and systems provided significant help in organizing the conference as well as providing some resources. We are grateful to them all. We express our gratitude to keynote, invited, and individual conference/tracks and tutorial speakers - the list of speakers appears on the conference web site. We would also like to thank the followings: UCMSS (Universal Conference Management Systems & Support, California, USA) for managing all aspects of the conference; Dr. Tim Field of APC for coordinating and managing the printing of the proceedings; and the staff of Monte Carlo Resort (Convention department) at Las Vegas for the professional service they provided. Last but not least, we would like to thank the Co-Editors of MSV’17: Prof. Hamid R. Arabnia, Prof. Leonidas Deligiannidis, and Prof. Fernando G. Tinetti. We present the proceedings of MSV’17. Steering Committee, 2017 http://americancse.org/
- 11. Contents SESSION: SIMULATION, TOOLS AND APPLICATIONS Generating Strongly Connected Random Graphs 3 Peter Maurer Simulating Virtual Memory Allocations using SPEC Tools in Microsoft Hyper-V Clouds 7 John M. Medellin, Lokesh Budhi Autonomously Battery Charging Tires For EVs Using Piezoelectric Phenomenon 13 Muhammad Kamran, Raziq Yaqub, Azzam ul Asar Traffic Re-Direction Simulation During a Road Disaster/Collapse on Toll Road 408 in Florida 19 Craig Tidwell SESSION: MODELING, VISUALIZATION AND NOVEL APPLICATIONS Performance Enhancement and Prediction Model of Concurrent Thread Execution in JVM 29 Khondker Shajadul Hasan A Stochastic Method for Structural Degradation Modeling 36 Peter Sawka, Sara Boyle, Jeremy Mange Increased Realism in Modeling and Simulation for Virtual Reality, Augmented Reality, and Immersive Environments 42 Jeffrey Wallace, Sara Kambouris A Toolbox versus a Tool - A Design Approach 49 Hans-Peter Bischof Generating Shapes and Colors using Cell Developmental Mechanism 55 Sezin Hwang, Moon-Ryul Jun Modeling Business Processes: Events and Compliance Rules 61 Sabah Al-Fedaghi Constructing a Takagi-Sugeno Fuzzy Model by a Fuzzy Data Shifter 68 Horng-Lin Shieh, Ying-Kuei Yang SESSION: NOVEL ALGORITHMS AND APPLICATIONS + IMAGING SCIENCE + SIGNAL ENHANCEMENT AND WIRELESS INFRASTRUCTURES Estimating Cost of Smart Community Wireless Platforms 75 Sakir Yucel
- 12. Detection of Ultra High Frequency Narrow Band Signal Using Nonuniform Sampling 82 Sung-won Park, Raksha Kestur Smart Community Wireless Platforms 87 Sakir Yucel Sketch Based Image Retrieval System Based on Block Histogram Matching 94 Kathy Khaing, Sai Maung Maung Zaw, Nyein Aye Smart City Wireless Platforms for Smart Cities 100 Sakir Yucel Measuring Benefits, Drawbacks and Risks of Smart Community Wireless Platforms 107 Sakir Yucel Magnetic Resonance Image Applied to 3-Dimensional Printing Utilizing Various Oils 114 Tyler Hartwig, Zeyu Huang, Sara Martinson, Ritchie Cai, Jeffrey Olafsen, Keith Schubert SESSION: POSTER PAPERS The Lithium-ion Battery Cell Analysis using Electrochemical-Thermal Coupled Model 121 Dongchan Lee, Keon-uk Kim, Chang-wan Kim Renfred-Okonigene Children Protection System Network: Where Is My Baby? 123 Dorcas Okonigene, Robert Okonigene, Clement Ojieabu
- 13. SESSION SIMULATION, TOOLS AND APPLICATIONS Chair(s) TBA Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 1 ISBN: 1-60132-465-0, CSREA Press ©
- 14. 2 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 15. Generating Strongly Connected Random Graphs Peter M. Maurer Dept. of Computer Science Baylor University Waco, Texas 76798 Abstract – The algorithm presented here is capable of generating strongly connected graphs in a single pass that requires O(n) time. The method is to create a spanning tree that is a directed acyclic graph, and adding a minimal number of edges to make the spanning tree strongly connected. This is done in a way that is completely general. Once the strongly connected spanning tree is created, additional edges can be added to the tree to create an arbitrary strongly connected graph. 1 Introduction One important problem in many types of simulation is creating random data for input to the simulation. Over the years, our need for such data in the simulation of gate-level circuits has led us to create a package for creating many different types of random data [1,2]. Despite its utility, this package has become dated. It was originally intended to generate random data files for input to a simulation program. Recently, we have begun a project to upgrade this package with new features to make it more useful for modern types of programs that do not depend heavily on file-based input, and to generate types of data that are more suitable to modern programs than character strings and simple binary values. One major focus of this activity (there are many) is the generation of random graphs. Graphs can be used to model many real-world phenomena. There are too many applications to mention individually, but see [3] for an example. One new feature of our package is the creation of graph-generation subroutines that can be incorporated into existing software. The graphs are generated internally as adjacency lists and passed, as pointers, to the simulation software. Graph specifications are simple, typically one line, but permit the specification of many different types of random graphs. The most common models are the edged-oriented models, the Gilbert model [4] and the Erdős–Rényi model [5]. The vertex-oriented models, power- law [6] and degree-sequence [7] models are also fairly common. This paper will focus on the edge-oriented models. The Gilbert model assigns a probability of p to the existence of any edge, and the Erdős– Rényi model assigns equal probability to all graphs with M edges. For the Gilbert model we generate k vertices and add each edge from the complete graph with probability p . For the Erdős–Rényi model, we sort all edges into random order and choose the first M edges from the sorted list. (The parameters k, i, and M are specified by the user.) The power-law and degree-sequence models are also available in our package, but these are beyond the scope of this paper. Parameters can be used to specify that the graph is directed, or that the graph must be connected, or both. Creating a connected non- directed graph is relatively simple. We generate a spanning tree using Algorithm 1, and then apply either the Gilbert or the Erdős–Rényi model to the remaining edges. The purpose of Algorithm 1 is to create a spanning tree for the graph. A non-directed graph is connected if and only if a spanning tree exists. Algorithm 1 adds Vertex 0 to the spanning tree, and then adds the remaining vertices by selecting a random vertex from the partial spanning tree as the parent of the new vertex. 1. Add Vertex 0 to the tree. 2. For each vertex, i, 1 through 1 k a. select a vertex j at random from the existing tree vertices b. Add an edge between i and j. c. Add vertex i to the tree. Algorithm 1. Creating the Spanning Tree. When the graph is directed, simply creating a spanning tree is insufficient because the resultant graph must be strongly connected. The spanning tree is still necessary, because there must be a path from Vertex 0 to every other vertex. When creating a spanning tree for a directed graph, the first step is to modify step 2 of Algorithm 1 so that the new edge proceeds from the tree vertex to the new vertex. This insures that there is a path from Vertex 0 to every other vertex. The resultant tree is a directed acyclic graph with the root of the tree as the only source. The sinks are the leaves of the tree. There are several straightforward methods for making the spanning tree strongly connected. When adding a tree vertex, we could add two edges, one from the tree vertex to the new vertex, and another in the opposite direction. This would make the tree strongly connected, but there would be no way to generate certain types of graphs such as the simple cycle of Figure 1. Another method is to add an edge from each leaf vertex to the root vertex. This is perhaps more general, but graphs such as that pictured in Figure 2 would be impossible to generate. Figure 1. A Simple Cycle. Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 3 ISBN: 1-60132-465-0, CSREA Press ©
- 16. Figure 2. Impossible for Leaf-to-Root. It is necessary to be able to generate any strongly connected graph on k vertices. Some attempts have been made to do so using the rejection method [8]. (The rejection method repeatedly generates directed graphs at random, and rejects those that are not strongly connected.) However, the rejection method works well only if the probability of rejection is small. When a sparse strongly connected graph is required, the rejection method is not suitable. Creation of the spanning tree is clearly the starting point for such a procedure, since it must exist. Furthermore, Algorithm 1 is able to generate any spanning tree, up to isomorphism. The next step must be to add additional edges to the graph to make it strongly connected. Furthermore, this should be done in a way that is certain to succeed and adds only a minimal number of edges to the tree. (The Leaf-to- Root method generates the absolute minimum number of edges, but is not suitable because it is not able to generate all strongly connected graphs.) To meet the needs of our generation software, we use a modified version of the Tarjan strongly connected component algorithm [9]. The Tarjan algorithm is capable of identifying any strongly connected graph by performing a single depth first search. The forward edges of the depth first search define a tree which is equivalent to our initial spanning tree. To detect a strongly connected graph, the Tarjan algorithm identifies a minimal set of back and cross edges to insure that the initial spanning tree strongly connected. Rather than detecting such edges (which do not exist in the initial spanning tree) we modify the algorithm to insert such edges where required. We do this in such a way that any suitable set of back or cross edge can be generated. Once the tree has been made strongly connected the Gilbert or Erdős– Rényi model can be applied to the remaining edges. 2 The Tarjan Algorithm To understand our method of inserting edges into the tree it is necessary to understand the principles of the Tarjan algorithm. The mechanism is based on depth first search with computed start times. Algorithm 2 shows the basic depth first search algorithm with the modification points that are used to detect strongly connected components. Start time is an integer in the range [0, 1] n where n is the number of vertices in the graph. The start time of vertex i is a sequential number indicating the order in which vertex i was first seen by the depth first search algorithm. In Algorithm 2, each element of the StartTime array (which is of size n) is initialized to 1 , and GlobalStartTime is initialized to zero. Both are global variables. 1. Function call DFS(i) 2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen 3. Increment GlobalStartTime by 1. 4. For each vertex j adjacent to vertex i a. If StartTime[j] is equal to 1 Call DFS(j) // back up to i 5. // backing up from i Algorithm 2, basic DFS. To detect strongly components, the basic DFS algorithm must be modified in three places. First, we add an array of size n named Low. After a vertex has been completely processed, it Low[i] will contain the smallest start time of any vertex that can be reached by following zero or more tree edges from Vertex i, followed by one back edge or cross edge. The following is added after step 2. 2.5. Set Low[i] equal to StartTime[i]. This step indicates that, initially, the lowest reachable start time is the start time of the current vertex. To step 4a we add Low[i] = min(Low[i],Low[j]). If it is possible to get further back than the current value of Low[i] by using tree edges, the new minimal start time recorded. The following step is added after step 4a: b. else if j is not already in a SCC, Low[i] = min(Low[i],StartTime[j]) If the edge (i,j) is a cross edge or a back edge, and it is further back than we have been able to get previously, its start time is recorded. The only problem that arises is when (i,j) is a cross edge to a strongly connected component that has already been identified. Such edges must be ignored. A status-array is normally used to keep track of such vertices. Finally, Step 5 is added to identify strongly connected components. If Low[i] is equal to StartTime[i], then there is no back edge or cross edge that provides a path from Vertex i to its parent. Therefore, Vertex i is in a different strongly connected component than its parent. 5. If Low[i] is equal to StartTime[i] Identify a new SCC. Tarjan’s algorithm also includes a stacking mechanism to identify the vertices belonging to a particular strongly connected component, but because our aim is to create a single strongly connected component, we will not consider this mechanism further. The same is true for the mechanism that tags vertices that have already been assigned to a strongly connected component. The full Tarjan algorithm is given in Algorithm 3. 4 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 17. 1. Function call DFS(i) 2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen 3. Set Low[i] equal to StartTime[i] 4. Increment GlobalStartTime by 1. 5. For each vertex j adjacent to vertex i a. If StartTime[j] is equal to 1 Call DFS(j), Low[i] = min(Low[i],Low[j]) b. Else if j is not already in a SCC, Low[i] = min(Low[i],StartTime[j]) 6. If Low[i] equals StartTime[i] Identify a new SCC Algorithm 3. The Full Tarjan Algorithm. 3 The modified Tarjan algorithm Our primary modification to the Tarjan algorithm is in Step 6, which identifies new strongly connected components. We wish to prevent this detection from taking place. In Step 6, the algorithm is backing up from Vertex i . All vertices with start times greater than or equal to StartTime[ i ] are in the subtree rooted at Vertex i . All vertices, j , that have start times less than i must meet one of the following three conditions. 1. Vertex j is an ancestor of Vertex i in the DFS tree. 2. Vertex j has already been assigned to a strongly connected component. 3. Vertex j and Vertex i have a common ancestor k , and k is reachable from j . Ignoring condition 2, it is clear that in Step 6, that Vertex i will be in the same strongly connected component as its parent if and only if there is an edge from a vertex v with a start time greater than or equal to StartTime[ i ] to a vertex w with a start time less than StartTime[ i ]. We modify Step 6 to insert such an edge when Low[i] is equal to StartTime[i]. Doing this also insures that condition 2 can never occur. Our modified algorithm is given in Algorithm 4. 1. Function call DFS(i) 2. Set StartTime[i] to GlobalStartTime. // vertex i is first seen 3. Set Low[i] equal to StartTime[i] 4. Increment GlobalStartTime by 1. 5. For each vertex j adjacent to vertex i a. If StartTime[j] is equal to 1 Call DFS(j), Low[i] = min(Low[i],Low[j]) b. Low[i] = min(Low[i],StartTime[j]) 6. If Low[i] equals StartTime[i] a. Set x to a random integer in the range [StartTime[i],GlobalStartTime] b. Set y to a random integer in the range [0,StartTime[i]-1] c. Identify the vertices v and w that correspond to the start times x and y. d. Add an edge from v to w. e. Set Low[i] equal to y. Algorithm 4. The Modified Tarjan Algorithm. Step 6c requires the inversion of the function that assigns start times to vertices. This is done in constant time by using an Inverse Start Time (IST) array and the following step following Step 2. 2.5 Set IST[StartTime[i]] to i In Step 6c, v and w correspond to IST[x] and IST[y], respectively. Step 6e negates the Low[i] equals StartTime[i] condition, since Vertex i is now in the same strongly connected component as its parent. The addition of the new edge may make the Low value for some of the other vertices in the tree rooted at Vertex i incorrect, but since these values will not be accessed after backing up from Vertex i, they do not need to be changed. Once Algorithm 4 has been run, the Gilbert or Erdős–Rényi models can be applied to add additional edges to the graph. The graphs created by Algorithms 1 and 4 are strongly connected and contain no more than 2 2 n edges. Algorithm 1 adds 1 n edges. Algorithm 4 can add at most one edge per vertex, and cannot add an edge to Vertex 0. Algorithm 4 can add edges in any possible way to make the tree strongly connected. When combined with the Gilbert or Erdős–Rényi models any strongly connected graph can be generated. Both Algorithm 1 and Algorithm 4 are O(n). The Gilbert or Erdős–Rényi models are both worst-case O( 2 n ) because they must consider all 2 2 n n potential edges. Although Algorithms 1 and 4 can generate any strongly connected tree up to isomorphism, there are many isomorphs that will never be generated. For example, Algorithm 1 always adds an edge from Vertex 0 to Vertex 1. If this is a problem, it can be solved by a random relabeling of the vertices after the graph is generated. 4 Experimental Results We ran several experiments to verify the effectiveness of our algorithm. The first experiment was to generate 10,000 5-vertex graphs, using the Gilbert model with 0 p , to verify that the examples of Figures 1 and 2 could be generated. The algorithm generated both these examples, along with many others. Figure 3 contains a sample of the generated graphs. Figure 3. Some Sample Graphs. Figure 3 demonstrates the essentially random nature of the generation process. Despite the fact that there is always an edge between Vertex 0 and Vertex 1, the structure of the graphs is obviously quite random, and they are obviously all strongly connected. This experiment was run using the Gilbert model with edge probability set to zero. This was done to show the structure of the strongly connected tree. Four other experiments were run to determine the performance of the algorithm. The hardware was an Intel 3.40 Ghz core I7-2600 with 4 cores and 8GB of memory, running Linux Red Hat version 3.10. A Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 5 ISBN: 1-60132-465-0, CSREA Press ©
- 18. single core was used to run the experiments. The Linux time command was used to obtain the timings. Algorithm 4 was implemented iteratively rather than recursively for increased performance. The results of the experiments are given in Figure 4. The first experiment generated one million five-vertex graphs to show the performance of the algorithm with small graphs. Because the time to generate a small graph is tiny, it was necessary to generate a large number of graphs so that the execution time would be measurable. The other three experiments were designed to show performance with large graphs. A single graph was generated because the time to generate such a graph is measurable. The Gilbert model was used for all four experiments (each edge having the same probability.) For the first experiment, we set the edge probability to .5, but it was necessary to set the edge probability to zero for the other three experiments due to memory requirements. If the edge probability were set to .5 for the one million vertex graph, half a terabyte of RAM (at least) would be required to store the graph. Experiment User time 1,000,000 5-vertex graphs Gilbert .5 p 3.8 seconds One 1,000,000-vertex graph, Gilbert 0 p .577 seconds One 10,000,000-vertex graph Gilbert 0 p 7.3 seconds One 100,000,000-vertex graph Gilbert 0 p 59.3 seconds Figure 4. Experimental Results. We speculate that it would take about 10 minutes to generate a one billion vertex graph, but 8GB of memory is insufficient to generate a graph of this size. 5 Conclusion The algorithm presented here is simple, easy to implement, and very fast. It can generate any strongly connected graph when used in conjunction with the Gilbert or Erdős–Rényi models, and possible node-relabeling. The algorithm should prove to be a useful tool for the generation of strongly connected graphs in most contexts. We have not yet addressed the vertex-oriented models, power- law and degree-sequence. Because each vertex has both an in-degree and an out-degree, it is not clear how to apply these models to directed graphs. It is necessary that the total of the in-degrees equal the total of the out-degrees. One model is to insist that the two degrees be identical for each vertex. Another model is to use the same set of degrees, but randomly distribute them over the vertices. It is also not clear whether the degree distributions should include the strongly connected tree edges, or whether these edges should be considered separately. For some degree distributions, it is not clear that a strongly connected graph even exists. We are currently working on these problems. 6 References 1. Maurer, P., “Generating Test Data with Enhanced Context-Free Grammars,” IEEE Software, Vol. 7, No. 4, July 1990, pp. 50-55. 2. Maurer, P., “The Design and Implementation of a Grammar- Based Data Generator,” Software Practice and Experience, Vol. 22, No. 3, March 1992, pp. 223-244. 3. Calvert, K., Doar, M., Zegura, E., “Modeling Internet topology,” IEEE Communications Magazine, Vol. 35, No. 6, June 1997, pp. 160-163. 4. Gilbert, E., (1959). “Random Graphs” Annals of Mathematical Statistics. Vol. 30, No. 4 1959, pp. 1141–1144. 5. Erdős, P.; Rényi, A., “On Random Graphs. I,” Publicationes Mathematicae, Vol. 6, 1959, pp. 290–297. 6. Aiello, W., Chung, F., Lu, L., “A Random Graph Model for Power Law Graphs,” Experimental Mathematics Vol. 10, No. 1, 2001, pp. 53-66. 7. Chatterjee, S., Diaconis, P., Sly, A., “Random Graphs with a Given Degree Sequence,” The Annals of Applied Probability, Vol. 21, No. 4, 2011, pp. 1400–1435. 8. Devroye, L, Non-Uniform Random Variate Generation, Springer-Verlag, New York, 1986. 9. Tarjan, R., “Depth-first search and linear graph algorithms,” SIAM Journal on Computing, Vol. 1, No. 2, 1972, pp. 146–160, 6 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 19. Simulating Virtual Memory Allocations Using SPEC Tools in Microsoft Hyper-V Clouds John M. Medellin*, Lokesh Budhi Associate Professor and Graduate Assistant at the Master of Science in Information Systems University of Mary-Hardin Baylor Belton, TX 76513-2599, USA Abstract— Private Clouds are gaining in popularity with small to medium sized businesses. By implementing a virtualized architecture, a company can gain strategic advantage through higher utilization of their technology assets. One of the first steps in determining if the virtualized architecture will make sense is to estimate the amount of resources required to actually create a private cloud. A risky approach is to take the applications that are executing and try to run them on a cloud. This option could turn out to be costly since key legacy applications are tightly coupled and in order to run the experiment one might need to move the entire system over (build the entire Cloud and port them). A second approach could be to model key parts of the system and test them with empirical models. This could also be costly and risky if key characteristics are erroneously estimated or omitted. Perhaps a better approach could be to use an industry simulation that can predict the usage patterns of similar systems and be configured to resemble workloads in production today. This paper executes simulations both on bare metal and within the Microsoft Cloud Stack (Windows 10, Windows Server 2012 R2 and Windows Hyper-V 2016) using the industry standard SPECjbb2015 simulation environment. We focus on measurement of incremental memory allocation and report throughput differences from two bare metal architectures (Windows 10 and Windows Server 2012 R2) to the target private cloud architecture. Our work begins by allocating 8GB to each environment and increases that variable to 10GB and 12GB. Significant performance gains are gained by increasing memory allocation in the virtual machine. We believe the contribution of this work is to demonstrate how industrial strength simulation tools can be applied to real world scenarios without having to completely build-out the architectures considered. This should be particularly useful to small companies that are contemplating private cloud implementations. Keywords— Hypervisors, Workload Simulation, Retail Applications, SPEC Corporation, Microsoft Windows 10, Microsoft Windows Server 2012 R2, Microsoft Hyper-V 2016 I. INTRODUCTION Clouds are used by many people and organizations today to gain a variety of advantages. There are many vendors and open sources for cloud software and an equal number of techniques for evaluating them. We can mix and match products that take advantage of our particular situation and the application workload profiles we are targeting. Each candidate architecture performs and enhances certain types of applications (referred to as “workloads”). The final selection will probably be based on the types of projected applications and their workload profiles [10]. Once we select the target Cloud tools they will also have to be tuned as far as memory allocations to deliver the expected results. Many studies have been published on the impact of resource virtualization and workload characteristics on Cloud architectures. Clouds essentially contain Virtual Machines and are managed by a central authority called a Virtual Machine Manager or a “Hypervisor”. Hypervisors can be secured from traditional vendors (e.g., Microsoft Hyper-V) or on open source like the OpenStack project [9]. Hypervisors are configured to either interact with the hardware directly (Type 1) or through a Host Operating system (Type 2) [3]. A typical implementation in smaller installations is the Microsoft Hyper-V, a type 2 architecture that runs on top of Windows Server 2012 R2 and manages virtual machines that can have Windows 10 or other guest operating systems. The applications themselves execute inside the virtual machines on the guest operating systems. A key strategy for measuring the performance of certain architecture attributes is the selection of an industry-standard simulation tool that will lend credibility to the results (e.g. it resembles what is to be measured). Simulation suites for measurement of a variety of attributes are provided by the Standards Performance Evaluation Corporation “SPEC”; www.spec.org. SPEC is a non-profit organization that was created by a consortia of major technology providers who have agreed on a set of principles to be used in building benchmarking tools. SPECjbb2015 is a simulated transaction generator that can provide for very complex scenarios in a retail grocery store. The system provides a set of simulation tools that can be applied to build a scalable model to resemble reality. When used to simulate Cloud performance, the tool will deliver compute transaction workloads (impacts on the Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 7 ISBN: 1-60132-465-0, CSREA Press ©
- 20. CPU/memory) that are executed in bare metal or on Virtual Machines (VMs). The functionality injects a series of standard transactions into a set of application processes that drive simulation through java applications provided. These transactions are progressively injected into the applications until the system gets saturated and can no longer provide sufficient throughput to keep up with the load being input (inputs exceed outputs). The fundamental objectives of this research are to first simulate the incremental overhead added by virtualization from bare metal to the Hyper-V environment and second to add more memory to determine the effects on overcoming the virtualization penalty. We use the SPECjbb2015 to simulate on bare metal using the Windows 10 operating system and the Windows Server 2012 R2 allocating 8GB in each case. Next, we virtualize that environment in a VM on Hyper-V, allocate the same amount of memory and execute the simulation. In a second experiment, we increase the allocated memory to 10GB and 12 GB and report the throughput statistics. The business applicability of the approach is discussed at the end of the document. In our analysis, we first present related work that has been done and how we have adapted some of those methods/results in our research. Next, we create a series of experiments which simulate the impact of virtualization as follows: • The SPECjbb2015 is executed on bare metal under the Windows 10 and the Windows Server 2012 R2 Operating Systems. Next, we virtualize the simulation under Windows Server 2012 R2 Network OS with the Hyper-V 2016 Hypervisor and Windows 10 guest operating system. All three of these have 8GB of Memory allocated. The results are reported in SPECjbb2015 throughput transaction totals. • The simulations above are repeated except under varied memory allocation at 10GB and 12GB. The corresponding increase throughput totals is reported. This research aims to demonstrate the usage of standard simulation tools in order to determine potential alternatives in Cloud resources without having to build the specific environments. The approach used could be scaled to other Cloud architectures than the one presented. II. RELATED WORK Virtualized environments date back a few decades. A key objective of virtualization was to keep the CPU busy while memory variables were being fetched from slower components in the computer [12]. With the advent of fully logically defined architectures in software (“software- defined systems”) we are now able to abstract the physical components into specifications resident in configuration files. The key software agent that manages and provisions the resources in a modern cloud is the Virtual Machine Monitor also referred to as the “hypervisor” [3]. All policies regarding allocation and usage of physical infrastructure resources are controlled by the hypervisor. Hypervisors are assisted by other tools and agents in order to deliver a fully functional Cloud Management Platform (CMP) [5]. A. Hypervisor Architecture Throughput In their review of open source hypervisors; Freet, Agrawal, Walker and Badr [5] detail out the general characteristics that give advantages of some over others. For example, their study includes adoption reviews on Eucalyptus, OpenStack, CloudStack, OpenNebula, Nimbus and Proxmox and presents a conclusion that OpenStack and CloudStack have over 30 times more messages in discussion forums that some of their other competitors (meaning they are more top of mind in the development community). They proceed to review the architecture fit within three commercial offerings (Xen, KVM, Virtual Box and ESX/VMware) in relation to the requirements for data center virtualization and infrastructure provision. In that study, various types of workloads are simulated through each candidate hypervisor and the throughput for each is reported. We have adopted a similar throughput reporting in our methodology. Vardhan Reddy and Rajamani [15] further study the incremental overhead added by 4 different hypervisors. Their work includes measurement of the residual CPU, memory, I/O (read/write) and network (send/receive) with focused workloads for Citrix XenServer, VMWare ESXi, Linux (Ubuntu) KVM and Microsoft Hyper-V. They conclude that the Hyper-V overall performance is very close to the winning VMWare. Their results are useful as another data point for our work (the work was done on a slightly older version than ours). In our opinion, the Microsoft architecture has continued to evolve in areas such as swap-file performance and such stack would perform at least as well as their findings indicate in similar tests. Their calculations on a 32GB cloud indicate that there is a 30% overhead on RAM at that level. Our experiments begin at 8GB memory allocation and they increment by 2 GB in successive trials until the system performance can be linearly approximated based on the increments. In yet a further diagnostic approach, Ye et. al. [16] propose a very innovative method and system for measuring usage of resources along the stack. They segment their findings into impacts on hardware (indicating cache optimization should be attempted), hypervisor (the overhead from the hypervisor itself) and finally from the virtual machines themselves (the workload profile). The Virt-B system reports the results from these layers as various workloads are being processed. This work not only quantifies the impacts on performance but further diagnoses the parts of the stack that might have significant bearing on the issue. B. Virtualization Overhead Optimization Virtualization of a platform’s resources can result in significant incremental requirements compared to bare metal 8 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 21. architectures. There are however, a set of tools and techniques that can help optimize those results in a virtualized environment. Oi and Nakajima [8] explored the effects of performance loads on the Xen. They determined that the performance of Xen could be enhanced in a virtualized environment by adjusting cache sizes in some applications in addition to incremental memory. Virtualized-Xen and bare- metal Linux were compared for throughput performance in different cache and memory optimization techniques. In most circumstances, it is a combination of both that will drive throughput gains in a virtual environment. In their work, they conclude that by varying configuration elements, a more effective use of resources can be achieved. The benchmarking system used was SPECjbb2001 and the effects of Network Interface Cards (NIC) were isolated so the workload could be measured in memory usage and throughput. Our team has adopted the SPEC performance suite as a workload simulator to determine the effects on memory allocations (another attribute) rather than in network throughput. Another relevant study is Jabry, Liu, Zhu and Panneerselvam [1]; hypervisor overhead impact is studied on the following resources: disk i/o, CPU, memory and VMM (hypervisor memory usage). The study predicts the usage of resources by the hypervisor in taking total resource usage and subtracting individual component loads and until only a residual is left (presumably the hypervisor load). Those tests were conducted with VMware, Virtual Box (Oracle Corporation) and Windows Virtual PC (Microsoft). Their work benchmarked a standard load in each hypervisor environment and used IOzone to quantify load on disk i/o, RAMSpeed to quantify the impact on memory and UnixBench, to indicate the effect on CPU. Their work concludes that the hypervisor is considerably higher on CPU rather than the other components of the architecture. Each suite of simulations focused on impacting a separate part of the architecture and demonstrated how different workloads impact the choice of hypervisor. It points to the Microsoft stack being more balanced due to its integration with the other components included in that specific Cloud architecture (MS Windows). We selected the Microsoft stack in our simulation so as to provide for greater integration between the components and being able to evaluate the environment as a “whole offering” from a single vendor. Further, tightness of coupling between the units would allow for study of the simulation as a whole without the need to study the effects of separate vendor “noise”. Chen, Patel, Shen and Zhou [2] studied virtualization overhead across multiple VMs running under Xen in cloud environments. They also found that the larger resource usage was attributable to the CPU. They also propose a series of equations that are remarkably accurate in predicting the lateral scaling of workloads on all components based on the observed results of the application under study. We provide a graphical analysis of throughput under several memory parameters (one of the parameters for optimization of CPU performance). C. Application Workload Research Based on the research referenced, there is a significant impact on utilization of CPU from the overhead generated by the hypervisor. Further the impact is based on the type of application that is operating in the virtualized environment. NasiriGerdeh, Hosseini, RahimiZadeh and AnaLoui [7] measured throughput degradation on Web applications using the Faban suite (a web-based workload generator). They simulated the behavior of heavy transactional Web applications that tend to be very network intensive. Their work also measured the effect on memory, disk i/o and CPU. They concluded that a disproportionate difference exists in CPU resources due to the translation of domain addresses. This work further confirms that the principal resource difference is the CPU utilization even when workloads may be more i/o bound (the penalties associated where in finding addresses; a CPU task, not access to the actual addresses in the Web environment; an i/o task). We incorporate this research by focusing on actual compute resource utilization rather than network or disk access. The SPECjbb2015 suite is focused on exhausting the compute resources rather than the disk (i/o) or network resources. San Wariya, Nair and Shiwani [11] focused their research on benchmarking three hypervisors; Windows Hyper-V, VMWare/ESXi and Citrix Xen in three cloud games; 3D Mark 11, Unigine Heaven and Halo. The objectives of their study are to identify which hypervisor was better from a cloud gaming workload perspective. The three performed differently in each category but were mostly lead by the VMWare product. For our purposes however, the HALO benchmark (number of frames per second) is probably the most predictive of workloads that are CPU bound. In this category, Hyper-V performed 7% ahead of VMWare and 57% ahead of Citrix Xen. This was another reason for selection of Hyper-V as the hypervisor for our test suite. D. The SPEC Benchmarking Suite The SPECjbb2015 constitutes a workload benchmarking simulation for a Supermarket Chain. The model can be extended to include several supermarkets and several central offices in a variety of virtual machine settings. The tool set can be configured in a variety of business transaction settings so that different business patterns can be simulated (e.g., web sales versus physical store sales). The system is owned and licensed by spec.org Error! Reference source not found. which is a consortium of major IT companies that have agreed on a set of principles to guide the performance benchmarking process. The system progressively injects transaction loads into the environment until saturation is reached. A sample output of these results is seen in Figure 3. In that graphic the system begins to stress at around the 5,200 java Operations Per Second (jOPS) with a range of 5K (median tolerance) to 50K Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 9 ISBN: 1-60132-465-0, CSREA Press ©
- 22. (max tolerance). The system reaches saturation (min tolerance) at around 6,700 jOPS and 60K. We report our results using the total transactions up to saturation. Figure 4 is a graphic representation of the architecture of the system. Figure 3: Sample SPECjbb2015 Benchmark Output www.spec.org Figure 4: SPECjbb2015 Architecture www.spec.org The SPECjbb products have been in existence since the late 1990s and are useful because of their industry acceptance. For example, Karlsson, Moore, Hagersten and Wood [6] used an earlier version (SPECjbb2001) and another application benchmark (ECPerf) to differentiate effects of cache misses between different types of applications. III. EXPERIMENT DESIGN As discussed above, experiments were designed where the same application (SPECjbb2015) was installed on: a) Bare metal with Windows 10 b) Bare metal with Windows Server 2012 R2 c) Virtual Machine: Windows Server 2012 R2 NOS / Hyper-V/ Windows 10 Guest OS The simulation was run for a typical store sales only company with 90% store sales and 10% online sales. This is typical of smaller stores that have not adapted to the online grocery demands of consumers and are experimenting with their own private clouds. B. Application Architecture Patterns The application patterns were analyzed by deriving use cases and preparing activity diagrams from the code for the application workload being simulated. i. Use Case Analysis Use cases are a functional decomposition tool that illustrate the process interactions between actors in applications [13]. The processes that we have selected in the SPECjbb2015 suite are fairly standard and follow similar patterns. The use case diagram for the store sales architecture is similar to this one (www.UML-diagrams.org), the “adornments” in the graphic describe the usage of artifacts. Figure 5: Store Use Case Diagram The inventory on-hand function at the physical store is susceptible to over-booked demand and out-of-stock conditions (where demand for an article exceeds supply). If the system detects an out-of-stock condition, it will proceed to cancel and back out the transaction. This process is memory intensive since it has to place the order items back into inventory and invalidate the order itself (see the error exception in the UML activity diagram below). ii. UML Activity Diagrams UML activity diagrams are a useful tool for analyzing the flow of logic through processes [13]. The following diagram was created from the code in the application. 10 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 23. Figure 6: Store UML Activity Diagram C. Bare Metal Implementations Two bare metal SPECjbb2015 implementations were used in the experiments; one on Windows 10 and one on Windows Server 2012 R2. The allocations were 8GB, 10GB and 12GB of memory (total of 6 bare metal simulations). D. Virtualization Hypervisor Architecture The third environment used MS Windows 10, Server 2012 R2 and Hyper-V [4]. The SPECjbb2015 software was compiled inside the virtual machine (VM). A full physical CPU, Network Interface Card (NIC) and all storage available was allocated to the VM. RAM of 8GB, 10GB and 12GB was allocated to the Virtual Machine. A diagram for this architecture is shown below, (ours has one VM). Figure 7: Virtualized Environment Architecture (www.microsoft.com) E. Infrastructure (Machine) Specifications The infrastructure environment that the experiments were executed on had the following specifications: • Hewlett-Packard Envy 15t • Intel i6700 quad-core processor • 16G RAM • 1TB Hybrid SSD • 4GB NVIDIA GTX 950M chip IV. EXPERIMENT RESULTS Figure 8 reports the total transaction throughput achieved under the different memory allocations in each environment. Figure 9 reports the percentage increase, using the 8GB results as the base for each environment. This is done to stress the incremental impact of each increase on the original base measurements. Figure 8: Throughput in the Experimental Systems The above indicates there were marginal increases in throughput on bare metal versus significant increases with increase in memory on the virtual machine. Figure 9: Percentage Gain in Throughput for the Experimental Systems The above emphasizes the gains throughput when memory is increased in the virtual machine. Throughput keeps increasing at significant rates (although begins to curb with the second increase in memory to 12GB). V. DISCUSSION & FUTURE PLANS The objectives of this study are to isolate the impact of additional memory allocation on a static workload. The hypothesis that additional memory increases throughput in virtualized environments. Part of this benefit is slowed as the allocation progresses. A. Results Discussion As systems are virtualized, they consume greater resources due to “virtualization overhead”; they require translation of the logical to the physical and back to the start Txi Sends SM an InStorePurchaseRequest Select a Random Customer Retrieve Customers Previous Purchase History Reserve Specific Quantity Of Each Product Calculate Total price Add the Available Discounts and Coupouns Customer Basket Validation Max Products Available Many Products to be Replenished Throw an Exception Proceed to Check out Generate Reciept Check Customer's Credit Customer has enough Credit Customer doesn't have enough Credit STOP Transaction Fails Move the Purchased Items from Store Inventory Debit the cost of Each Item from Customer's Account Send Suppliers a Request, If any item runs out from store Send Receipt back to HQ Stop Instore Activity 10GB 12GB Win 10 2% 2% Win 2012 20% 7% Win Hyp-V 95% 79% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Percent Throughput Improvement, 8GB Constant Base 10GB 12GB Win 10 2% 2% Win 2012 20% 7% Win Hyp-V 195% 37% 0% 50% 100% 150% 200% 250% Percent Throughput Improvement, 8GB Constant Base Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 11 ISBN: 1-60132-465-0, CSREA Press ©
- 24. logical. Our study has illustrated how using a process of simulation, a small company may avoid the costly and risky process of making the decision to virtualize in a private cloud without knowing how much incremental resource will be needed. We used Windows 10 and Server 2012 R2 to test on bare metal. It is important to understand that these are two fundamentally different Operating Systems. Windows 10 is focused on managing the desktop and execution of localized workloads. It provides rich functionality in areas such as graphics and gaming which are not simulated by the SPECjbb2015 suite but are nevertheless instantiated in it’s services. The Server 2012 R2 is a Distributed OS whose focus is to manage standard workloads associated with raw compute and storage power. Under these circumstances, the Server 2012 R2 performs with better results in traditional business process simulations like SPECjbb2015. The Hyper-V extension of the Windows Server 2012 R2 is a tool for managing the life cycle of virtual machines. It is an extension of the Network OS that communicates decisions to the virtualization layer for translation of operating parameters back and forth. This additional load constitutes overhead (more resources). Some of these resources are “fixed”; they are there by virtue of instantiation and some are “variable”; by usage of the workspace through applications. As memory is increased, there exists more workspace for applications and the overall impact of the hypervisor usage of memory is reduced. The first incremental memory allocation (from 8GB to 10GB) has a higher yield because a greater percentage of that “boost” goes to the application. The successive increment (from 10GB to 12GB) is still significant but not as high. According to hypervisor vendors and Reddy and Rajamani [15] this reduction will continue until the bare metal and virtualized environment will start to resemble each other in throughput given volume/memory mix. The simulation process allows a company to make plans of how to deploy in the future. Using a standard simulation may lead to answering some key questions as: a) Should we virtualize or keep on bare metal? b) When should we revisit our decision? A company could continue the simulation by contracting additional capacity with one of the major Cloud vendors and determine where the VM/bare metal results ultimately blur. B. Future Plans The team is busy executing additional work in running additional simulations that can implement optimization techniques. The ultimate objective is to have a “cookbook” of simulation/optimization techniques that can be used in private or hybrid cloud evaluators. REFERENCES [1] H. Al Jabry; L. Liu, Y. Zhu, J. Panneerselvam: “A Critical Evaluation of the Performance of Virtualization Technologies”. 9th International Conference on Communications and Networking in China (2014). p. 606-611. [2] L. Chen; S. Patel; H. Shen; Z. Zhou: “Profiling and Understaning Virtualization Overhead in Cloud”. 2015 44th International Conference on Parallel Processing. p. 31-40. [3] T. Erl; Z. Mahmood; R. Puttini: “Cloud Computing Concepts, Technology & Architecture”. c. 2013 Arcitura Education, Inc./Pearson Education, Upper Saddle River, NJ. USA. [4] A. Finn; M. Luescher; P. Lownds; D. Flynn: “Windows Server 2012 Hyper-V; Installation and Configuration Guide”. c. 2013 Wiley and Sons, Indianapolis, IN. USA. [5] D. Freet; R. Agrawal; J. Walker; Y. Badr: “Open source cloud management platforms and hypervisor technologies: A review and comparison”. SoutheastCon 2016. p. 1-8. [6] M. Karlsson; K.E. Moore; E. Hagersten; D.A. Wood: “Memory system Behavior of Java-Based Middleware”. The Ninth International Symposium on High- Performance Computer Architecture, 2003. HPCA-9 2003 p. 217-228. [7] R. NasiriGerdeh; N. Hosseini; K. RahimiZadeh; M. AnaLoui: “Performance Analysis of Web Application in Xen-based Virtualized Environment”. 2015 5th International Conference on Computer and Knowledge Engineering (ICCKE). p.258-261. [8] H. Oi; F. Nakajima: “Performance Analysis of Large Receive Offload in a Xen Virtualized System”. 2009 International Conference on Computer Engineering and Technology. p. 475-480 [9] J. Rhoton; J. De Clercq; F. Novak: “OpenStack Cloud Computing Architecture Guide 2014 Edition”. c. 2014 Recursive Press, USA & UK. [10] A. Salam; Z. Gilani; S. Ul Haq: “Deploying and Managing a Cloud Infrastructure”. c. 2014 Sybex, a Wiley Brand, Indianapolis, IN. USA [11] A. SanWariya; R. Nair; S. Shiwani: “Analyzing Processing Overhead of Type-0 Hypervisor for Cloud Gaming”. 2016 International Conference on Advances in Computing, Communication, & Automation (ICACCA) (Spring). p. 1-5. [12] W. Stallings: “Operating Systems: Internals and Design Principles 7th ed”. c. 2012 Pearson/Prentice Hall, Upper Saddle River, NJ. USA [13] C. Larman; “Applying UML and Patterns: An Introduction to Object-Oriented Analysis and Design and Iterative Development (3rd Edition)” c. Pearson Education 2005, Upper River, NJ. [14] www.spec.org [15] P.V. Vardhan Reddy; L. Rajamani: “Virtualization Overhead Findings of Four Hypervisors in the CloudStack with SIGAR”. 2014 World Congress on Information and Communication Technologies (WICT 2014) p. 140-145. [16] K. Ye; Z. Wu; B. Zhou; X. Jiang; C. Wang; A. Zomaya: “Virt-B: Toward Performance Benchmarking of Virtual Machine Systems”. IEEE Internet Computing, V. 18, Issue 3 (2014). p. 64-72 12 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 25. Autonomously Battery Charging Tires For EVs Using Piezoelectric Phenomenon Muhammad Kamran1 , Dr. Raziq Yaqub2 , Dr. Azzam ul Asar1 1 CECOS University, Peshawar, Pakistan, 2 University of Tennessee, USA Abstract:-This paper illustrates the use of piezoelectric material to generate electricity in the Electric Vehicle (EV) tire. According to the proposed mechanism, the vehicle tires are embedded with layers of the Piezo-electric material along the periphery. Thus, when the EV is in motion, electricity can be generated due to mechanical stress in that part of Piezo-electric material that is in contact with the road surface. The results show that, with peripheral arrangement of Piezo-material inside the automobile tire, we can generate electricity that can be stored in a battery to run the EV for some extra miles of the total miles the EV is capable of on a single charge. The use of Polyvinylidene fluoride (PVDF); a polymer based piezoelectric material is considered due to its robust and favorable properties. Keywords— Piezoelectricity, Polyvinylidene fluoride (PVDF), Electrical power, Mechanical stress-to-electricity conversion, Automobile tire. 1 Introduction Due to the rising demand for generating energy in the most efficient way; smart, intelligent and adaptive materials are being used and one such smart substance is the piezoelectric material. Piezoelectric substances produce electric charge when mechanical stress is applied on its surface. Piezoelectric materials are composed of various materials namely crystals, ceramics, polymers etc. Polymer- based piezoelectric materials have served as the most efficient material compared to ceramics and crystals for applications where elasticity is preferred. The most commonly used polymer based piezoelectric material is Polyvinylidene Fluoride (PVDF). PVDF is a transparent, semi-crystalline, thermoplastic fluoroplastic. We have employed PVDF as the piezo electric material in our work based on the merits of PVDF as listed below [13]: (i) Piezoelectricity obtained from PVDF is several times greater than that obtained from quartz or ceramics. (ii) PVDF materials are insoluble in water, resistant to solvents, acids, bases, heat, and generate low smoke in case of any fire accidents. (iii) Has low weight and low thermal conductivity. (iv) Highly resistant to chemical corrosion and heat variations, thus withstands exposure to harsh chemical and thermal conditions. (v) Very good mechanical strength and toughness and has high abrasion resistance. (vi ) Low permeability to most gases and liquids. (vii) Unaffected by long-term exposure to ultraviolet radiation. (viii) Less expensive compared to its counterpart. These features make them most suited to be employed in EV tires. However, proof of concept needs to be done, that requires collaboration with tires manufacturer. The rest of the paper is divided into the following sections. Section-2 describes the composition and structure of the tire, section-3 explians our proposal on embedding piezo-electric material in tire and harvesting energy from it. Section-4 calculates cost efficiency of the proposed mechanism, Section-5 suggest future work, Section-6 concludes the paper, Section-7 lists some of the key references, and finally Section-8 titled as Annex, presents the detailed mathematical analysis. Due to the rising demand for generating energy in the most efficient way; smart, intelligent and adaptive materials are being used and one such smart substance is the piezoelectric material. Piezoelectric substances produce electric charge when mechanical stress is applied on its surface. Piezoelectric materials are composed of various materials namely crystals, ceramics, polymers etc. Polymer- based piezoelectric materials have served as the most efficient material compared to ceramics and crystals for applications where elasticity is preferred. The most commonly used polymer based piezoelectric material is Polyvinylidene Fluoride (PVDF). PVDF is a transparent, semi-crystalline, thermoplastic fluoroplastic. We have employed PVDF as the piezo electric material in our work based on the merits of PVDF as listed below [13] 2 Composition and Structure of the Tire The most basic component in the tire is “rubber” which may be ‘synthetic rubber’ or ‘natural rubber’. Other components that are present in the tire are fabric wire, polymers, fillers, softeners, anti-degradents and curatives. As polymers are the backbone of rubber compounds, it is more appropriate to embed polymer piezoelectric material within the structure of the tire as done in [1]. Since the objective of using Piezoelectric material in [1] is sensing, it simply employs pallets of PVDF materials. However, we embed PVDF material as a circular ring along the entire periphery of the tire to maximize electricity generation. There are 3 main categories of tires such as Diagonal (bias) tire, belted bias tire and Radial tire. Radial tires are most commonly used in the automobile industry; therefore, this paper considers radial tire for mathematical analysis. However, it does not preclude other types. Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 13 ISBN: 1-60132-465-0, CSREA Press ©
- 26. Fig. 2 Piezoelectric Material as a Circular Ring in a radial tire 3 Proposal on Embedding Piezoelectric Material in Tires to Harvest Energy Figure 1 shows the overall concept of piezoelectric generation phenomenon in context to the proposed tire scenario. Fig. 1. Overall Concept of Piezoelectric Generation in Proposed Tire. Figure 2 shows the cross-sectional area of the original radial tire, where we proposed to have one layer of piezoelectric material, along the periphery of the tire, below the rubber layer or any suitable place, the tire manufacturers deem suitable. Having the PVDF material as a circular ring along the entire periphery of the tire is considered to be more efficient compared to having pallets of PVDF material embed within the tire. It is because when the tires will rotate, most portions of the PVDF will be coming in contact with the road, and thus will ensure constant generation of electricity. Electricity is generated in piezoelectric materials due to mechanical stress in that part of piezoelectric material that is in contact with the road surface. Piezoelectricity is the direct result of the piezoelectric effect. The electricity so produced is fed to the car battery. Having the PVDF material as a circular ring along the entire periphery of the tire is more efficient compared to having pallets of PVDF material embed within the tire. When the tires will rotate at a high speed, most portions of the PVDF will be coming in contact with the road, and thus will ensure constant generation of electricity as shown in figure 1. In this section, pressure exerted by the road on the automobile tire is modeled and the amount of energy harvested in this process is calculated. Since the section involves mathematical variables, therefore for the convenience of the readers, the terminology, abbreviations and units are provided in the form of a table below: TABLE I. NOMENCLATURE, SYMBOLS AND UNITS Pa ra- meter Parameter definition Unit A Area meter2 c Circumference centimeter C Charge Coulomb d Piezoelectric strain coefficient or D Diameter centimeter F Force Newton g Gravity meter/sec2 I Current Ampere k Distance centimeter m Mass of the car kilogram Pp Charge surface density Coulomb/m2 p Power Watts P Pressure Newton/m2 t1 Time seconds T Pressure exerted on the PVDF material Newton/met er2 v Velocity/speed Miles/hour V Voltage volts W Width Centimeter g Appropriate piezoelectric coefficient for the axis of applied stress or strain or t Thickness of a ring µm Modeling the System The experiments performed by the Curie brothers demonstrated that the Charge Surface Density is proportional to the pressure exerted, and is given by [2] Pressure exerted by a car on the road can be given by, Tire With Piezoelectric Material Road Surface Piezoelectric Phenomenon Tire With Piezoelectric Material Road Surface Piezoelectric Phenomenon 14 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 27. Where F is the force exerted on the tire and is equal to the weight (F = mg, where m is the mass of the car and g is the force of gravity and is 9.8 m/s2 ). And A is the tire surface area in contact with the road. The tire surface area A in contact with the road can be calculated as A = π x D x W x 0.1 Where 0.1 is due to the fact that 10% of the tire area is in contact with the surface of the road [3]. Output voltage for the given stress or strain is given by V0 = g3n x Xn x t Where n = 1, 2 or 3 and Xn = T [7]. As we are considering n=1 i.e. (in the piezoelectric material the electrical axis is always fixed as it is three in this case and the mechanical axis is either one, two or three), the value of g31 is specified in table 2. Moreover we are considering force on tire due to weight along Y direction as shown in figure. The component of weight acting in this way is constant and 20% of total weight [14] [15]. Now if the force is assumed to be acting axially then the area should also be taken in the specified direction. So the V0 will be modified as [7]. V0 = g31 × (Force/width × thickness) × thickness = g31 × (Force/width) Fig 3. [15] Fig. 4 [15] Fig. 5. [15] Result Calculation If the mass of the car is 1500Kg (Because electric vehicles have more weight due to their battery) then the calculated force is 14700Newtons (3675N for each tire). When the average diameter of the PVDF ring is considered to be D = 0.5588 m (22 inch), width of a ring is 0.1651 m (6.5 inch) and thickness of the ring is 110 micrometer [7]. Therefore, the Area (A) for the PVDF ring is 0.289 m2 and the surface area when 10% of the tire is in contact with the road surface is 28.9 × 10-3 m2 . From this value of force (F) and area (A) the pressure (T) exerted on the PVDF material is 127162.63 N/m2. Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 15 ISBN: 1-60132-465-0, CSREA Press ©
- 28. TABLE II. TYPICAL PROPERTIES OF PVDF [7] Parameter Sy mbol Valu e Unit Thickness t 55 µm (micron, 10-6 ) Piezoelectric Strain Constant d31 23 10-12 or Piezoelectric Stress Constant g31 216 10-3 or According to above PVDF properties and equation the charge surface density is 2.92 × 10-6 C/m2 . We know that the, Charge = Current x Time, and Power = Current x Voltage Assumption has been made in order to make calculations easier, that the amount of time taken to generate the electricity is one second, allowing the charge to be equal to the current [3]. From all of the above equations and assumptions the power generated is 2.81mW for 1 tire. We consider only 10% of the tire touches the road surface, so in one rotation 10 times electricity is generated. In one rotation 28.1mW power is generated. The US environment protection agency official range is 117 km (73 mile) with an energy consumption of 765 kilojoules per kilometer or 34 KWh/100mile [6] [12]. According to this energy consumption our designed car can run extra 39 km because the power generated by four tires of the car with any specified speed. 4 Cost Efficiency Cost of the tire after incorporating the PVDF ring inside the tire depends mainly on two factors, cost of the PVDF and the cost for implementing the PVDF ring inside the tire. The PVDF ring cost is from $50 - $100 per Cubic Meter, which depends on length, width and thickness of the ring [4]. If we consider the cost of the PVDF to be used in four tires to be $50 including embedding process of PVDF material inside the tire. The average life of all-season radial tire advertised by the manufacturer is 50000 miles. Using proposed technology, the EV can bring a cost saving worth 17500 additional miles. If the cost of electrical energy is $0.04/mile, a saving of $525 can be achieved. 5 Future Work We plan to do the following work in future: Followings tasks would be carried throughout the length of the project. (i) Produce simulations considering the tire industry standards such as tread, the body with sidewalls, and the beads (the rubber-covered, metal-wire that hold the tire on the wheel) (ii) Use Autodesk (It is already licensed to UT) sofware to simulate the performance of design parameters such as distribution of PVDF around the periphery of the tire. (iii) Simulate the effects of different typs/concentrations of PVDF compounds and different types of distribution of PVDF around the periphery of the tire. And also using different concentrations of PVDF in different parts of the tire. (iv) Simulate the effects of different typs/concentrations of rubber compounds and different types of distribution of rubber around the periphery of the tire. And also using different concentrations/ratios of rubber and PVDF in different parts of the tire. (v) Analyze the effects of different stresses on the proosed tire design (emulating different weights, road roughness, etc.), and discover design limitations. (vi) Analyze the effect of different temperatures (eulating different hot/cold weathers). Also simulate the effects of different sizes the tires come in. It has to be ensured that simulations as well as timelines for simulations meet the expectations, through validation, and comparison with the specifications of standard tires (non PVDF tires). 6 Conclusion Our work demonstrates a method of generating electricity using the PVDF material. Mathematical analysis proves that the EV can run extra 37 miles on a single charge with a speed of the car is 60mph. Since the cost of PVDF and its implementation is not so expensive, a saving of about $500 is expected over the life of the tire. Overall, the proposed method is an excellent choice to generate power when the car is on move. Acknowledgment The 1st and 3rd authors would like to acknowledge the technical support of Dr. Raziq Yaqub for his valuable contribution extended during the course of this project and allowing to improve the mathematical model. References [1] Jingang Yi, "A Piezo-Sensor-Based 'Smart Tire' System for Mobile Robots and Vehicles",March 2007. [2] Arnau, Antonio. “Fundamentals on Piezoelectricity.” Piezoelectric Tranducers and Applications. New York, 2008. Print, pp.4. [3] http://cosmos.ucdavis.edu/archives/2011/cluster2/Yau_ Derek.pdf. [4] http://www.alibaba.com/product- gs/322181211/PVDF_Intalox_Saddle_Ring.html [5] http://www.carfolio.com/specifications/models/car/car= 107844&GM. 16 Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | ISBN: 1-60132-465-0, CSREA Press ©
- 29. [6] http://www.autoblog.com/2009/08/01/2010-nissan-leaf- electric-car-in-person-in-depth-and-u-s-b/ [7] Measurement Specialities Inc. April 1999 “Piezo Film Sensor Technical Manual”, P/N 1005663-1, Rev. B, pp. 3-4,28 [8] West Coast Green Highway. Electric Highways Project, 2010. Retrieved March 09, 2012, from http://west coastgreenhighway.com/electrichighways.htm [9] Kris De Decker, May 3, 2010. “The status quo of electric cars: better batteries, same range.” Low-tech magazine. Retrieved March 13, 2012 from http://www.energybulletin.net/node/52736. [10] http://www.tirerack.com/tires/tiretech/techpage.jsp?tec hid=46 [11] http://www.tirerack.com/tires/tiretech/techpage.jsp?tec hid=7. [12] http://teacher.pas.rochester.edu/phy121/lecturenotes/Ch apter06/Chapter6.html. [13] http://www.openmusiclabs.com/wp/wp- content/uploads/2011/11/piezo.pdf [14] http://en.wikipedia.org/wiki/Nissan_Leaf [15] http://en.wikipedia.org/wiki/Polyvinylidene_fluoride [16] http://www.mate.tue.nl/mate/pdfs/8351.pdf [17] http://road-transport- technology.org/Proceedings/2%20- %20ISHVWD/Vol%201/TRUCK%20TIRE%20TYPE S%20AND%20ROAD%20CONTACT%20PRESSUR ES%20-%20Yap%20.pdf [18] http://www.tzlee.com/blog/?m=201103 Annex: Detailed Mathematical Analysis Calculating Patch Area of PVDF Ring Contact patch (also called footprint) is the area in which the tire is in contact with the road surface). Different vehicles have different contact patch depending on tire’s diameter and width. Tires diameter ranging from 8 to 26 which are given in detail in [8] [9]. For the sake of analysis we considered a tire with the diameter of 22 inches and width of 6.5. (I.e. the tire size of 185/55R15 commonly used for Passenger Electric Vehicles (EV)). We consider incorporating PVDF ring inside the whole width of tire, so that the tire continue to adhere with its original texture without scarifying its original purpose or violating its specifications in terms of road resistance, air pressure, etc. D = 0.5588 meters W = 0.1651 meters Therefore, Area (A) is given by, ATotal = ∏ × D × W = ∏ × 0.5588 × 0.1651 = 0.289 m2 We assume that only 10% of the area is in contact with the road surface. Therefore, the Area (A) is given by, A = 28.9 × 10-3 m2 To determine the Pressure If we consider the mass of the car to be 1500 kg, force can be calculated as follows: Force = m × g = 1500 × 9.8 = 14700 N And Force on one tire will be = 14700/4 =3675N Therefore Pressure = 14700 / 28.9 × 10-3 = 127162.63 N / m2 To determine the charge surface density The charge surface density is given by, Where d is the piezoelectric strain coefficient and from table 2, its value is given to be 23 × 10-12. Therefore, = 23 × 10-12 × 127162.63 = 2.92 × 10-6 C/m2 To determine the output voltage In the previous section, we have briefly discussed the equations for Output voltage which is given by, V0 = g3n x Xn x t In our work we consider n = 1 and the value of g31 and t is specified in table 2. Xn = (Force/width × thickness) Therefore, output voltage is, V0 = g31 × (Force/width × thickness) × thickness = g31 × (Force/width) = 216 × 10-3 × (0.2×3675/0.1651) = 961.6 V And 0.2 is the component of force that acts axially shown in figure-2c, as we are considering g31 mode so the force should also be considered along the specified direction. To determine the Total Power Total Power = Charge Surface Density x Output voltage = 2.92 × 10-6 × 961.6 = 2.81 mW We consider only 10% of the tire touches the road surface, so in one rotation 10 times electricity is generated. In one rotation of one tire 28.1mW power is produced. Int'l Conf. Modeling, Sim. and Vis. Methods | MSV'17 | 17 ISBN: 1-60132-465-0, CSREA Press ©
- 30. Discovering Diverse Content Through Random Scribd Documents
- 31. O. Goth. wari, a mountain. Garritour, s. The watchman on the battlements of a castle. K. Hart. GARRON, GERRON, s. 1. A small horse, S. Ir. id. a hackney. Stat. Acc. 2. An old stiff horse, Loth. 3. A tall stout fellow, Ang. Ir. garran, a strong horse. GARRON NAILS, Spike nails, S. GARSON, s. An attendant. Sir Gawan. Fr. garçon, a boy. GARSTY, s. The resemblance of an old dike, Orkn. Isl. gardsto, locus sepimenti. GARSUMMER, s. Gossamer. Watson. GART, GERT, pret. of Gar, Ger. GARTANE, s. A garter, S. Chron. S. P.
- 32. Gael. gairtein, id. GARTEN BERRIES, Bramble berries. Gl. Sibb. GARTH, s. 1. An inclosure. Wallace. 2. A garden. Dunbar. A. S. geard, used in both senses. GARVIE, s. The sprat, a fish, S. Garvock, Inverness. Sibbald. To GASH, v. n. 1. To talk a great deal in a confident way, S. 2. To talk pertly, or insolently, S. 3. To talk freely and fluently, S. synon, gab. Burns. Fr. gauss-er, to gibe. Roquefort gives O. Fr. gas, gaz, as merely a variation of gab, plaisanterie, moquerie. Gash, s. 1. Prattle, S. synon. gab. 2. Pert language, S. Gash, adj. 1. Shrewd in conversation, sagacious, S. Watson. 2. Lively and fluent in discourse, S. Ramsay. 3. Having the appearance of sagacity conjoined with that of self-
- 33. importance, S. Burns. 4. Trim, respectably dressed, S. R. Galloway. GASH, s. A projection of the under jaw, S. To Gash, v. n. 1. To project the under jaw, S. 2. To distort the mouth in contempt, S. Fr. gauche, awry; gauch-ir, to writhe. GAST, s. A gust of wind, S. B. A. S. gest, id. GASTROUS, adj. Monstrous, Dumfr. Dan. gaster, Manes, ghosts. O. E. gaster, to affright. GATE, s. A way. V. Gait. GATE, s. Jet. V. Get. Douglas. GATING, s. Perhaps, guessing. Burel. Su. G. gaet-a, conjecturam facere.
- 34. GAUCY, GAWSY, adj. 1. Plump, jolly, S. Journal Lond. 2. Applied to any thing large, S. Burns. 3. Metaph., stately, portly, S. Ferguson. Su. G. gaase, a male. The ancient Gauls called strong men Gaesi. 4. Well prepared, S. A. Douglas. GAUCKIT, adj. Stupid. V. Gowkit. GAUD, GAWD, s. 1. A trick. Douglas. 2. A bad custom or habit, S. B. Fr. gaud-ir, to be frolicksome, Su. G. gaed-as, laetari; from Isl. gaa, gaudium. GAVEL, GAWIL, s. The gable of a house, S. Wyntown. Su. G. gafwel, Belg. gevel, id. GAVELOCK, s. An iron lever, S. A. S. gafelucas, hastilia, gafl, furca.
- 35. GAUGES, s. pl. Wages. Acts Sedt. O. Fr. guaige. GAUKIE, GAWKY, s. A foolish person. V. Gowk. Sw. gack, id. Ramsay. Gaukit, Gawkit, adj. Foolish, giddy, S. Morison. GAUL, s. Dutch myrtle, S. V. Scotchgale. GAULE, s. A loud laugh. V. Gawf. GAUT, s. A hog, S. Sir J. Sinclair. Isl. galt, sus exsectus. To GAW, v. a. 1. To gall, S. Ferguson. 2. Metaph., to fret, S. Ramsay. To Gaw, v. n. To become pettish, Loth. Ramsay.
- 36. Gaw, s. The mark left by a stroke or pressure, S. Polwart. GAW, s. A gall-nut. Ramsay. GAW, s. 1. A furrow or drain, S. Statist. Acc. 2. A hollow with water springing in it, Ang. GAWD, s. A goad, S. Ross. GAWDNIE, GOWDNIE, s. The yellow gurnard, S. q. gold-fish. Sibbald. To GAWF, GAFF, v. n. To laugh violently, S. Ramsay. Su. G. gaffla, id. Germ. gaffen, to gape. Gaulf, Gawf, Gaffaw, A horse-laugh, S. Knox. To GAWP UP, v. a. To swallow voraciously, S. Ramsay. Sw. gulpa, buccis vorare deductis. Gawp, s. A large mouthful, S.
- 37. GAWRIE, s. The red gurnard, S. Sibbald. GAWSIE, adj. Jolly. V. Gaucy. GEAN, GEEN, s. A wild cherry, S. Fr. guigne, guine, id. Statist. Acc. Geantree, s. A wild cherry-tree, S. Statist. Acc. GEAR, GEARED. V. Gere. GEARKING, part. adj. Vain. Lyndsay. A. S. gearc-ian, apparare. GEAT, s. A child. V. Get. GEBBIE, GABBIE, s. The crop of a fowl, S. Ferguson. Gael. ciaban, the gizzard. To GECK, GEKK, v. a. 1. To sport, Ang. 2. To deride, S.
- 38. Philotus. 3. To befool. Leg. St Androis. 4. To jilt, S. 5. To toss the head disdainfully, S. Ramsay. Teut. gheck-en, deridere, Su. G. geck-as, ludificari, Sw. gaeck-a, to jilt. Geck, Gekk, s. 1. A sign of derision. Dunbar. 2. A jibe. Montgomerie. Teut. geck, jocus. 3. Cheat, S. Poems 16th Cent. GED, (g hard) s. The pike, a fish, S. Su. G. Isl. gaedda, id. Barbour. Ged-staff, s. 1. A staff for stirring pikes from under the banks. Douglas. 2. A pointed staff, from Su. G. gadd, aculeus. Gl. Sibb. GEE, (g hard) s. To tak the gee, to become pettish and unmanageable, S. Isl. geig, offensa. Ross.
- 39. GEY, GAY, (g hard) adj. Tolerable. S. P. Repr. A gey wheen, a considerable number. Gey, Gay, adv. Indifferently. Gey and weil, pretty well, S. Ramsay. Geily, Gayly, Geylies, adv. Pretty well, S. Kelly. Teut. gheef, sanus; Su. G. gef, usualis. GEYELER, s. Jailor. Wallace. To GEIF, GEYFF, v. a. To give. Douglas. To GEIG, (g soft) v. n. To make a creaking noise, S. Douglas. Germ. geig-en, fricare. GEIG, s. A net used for catching the razor-fish. Evergreen. GEIL, GEILL, s. Jelly, S. Fr. gel. Lyndsay. GEILL POKKIS, bags through which calfshead jelly is strained. Maitland P.
- 40. GEING, (g hard) s. Intoxicating liquor of any kind, Ang. Isl. gengd, cerevisiae motus. GEING, (g hard) s. Dung, Bord. A. S. geng, latrina. GEIR, s. Accoutrements, &c. V. Ger. To GEYZE, GEISIN, GIZZEN, (g hard) v. n. To become leaky for want of moisture, S. Ferguson. Su. G. gistn-a, gisn-a, id. GEIST, s. 1. An exploit; Lat. gesta. Douglas. 2. The history of any memorable action. Douglas. GEIST, GEST, s. 1. A joist, S. Doug. 2. A beam. Barbour. GELORE, GALORE, GILORE, s. Plenty, S. Gael. go leoir, enough. Ross.
- 41. To GELL, (g hard) v. n. To thrill with pain, S. Sir Egeir. Germ. gell-en, to tingle. To GELL, (g hard) v. n. To crack in consequence of heat, S. Isl. geil, fissura. Gell, s. A crack or rent in wood, S. GELL, (g hard) s. A leech, S. B. gellie, Perths. Su. G. igel, id. C. B. gel, a horseleech. GELT, s. Money. V. Gilt. GEN, prep. Against. A. S. gean, id. GEND, (g hard) adj. Playful. S. P. Repr. Isl. gant-a, ludificare. GENYIE, s. Engine of war. Minst. Bord. GENYEILD, GENYELL, s. V. Ganyeild.
- 42. GENIS, s. Apparently, the rack. Act Sed. Fr. gêne, id. from Lat. gehenna. GENYUS CHALMER, bridal chamber. Douglas. GENTY, (g soft) adj. Neat, limber, elegantly formed, S. Ramsay. Teut. jent, bellus, elegans. GENTIL, adj. Belonging to a nation. Douglas. GENTILLY, adv. Completely, Ang. Barbour. GENTRICE, GENTREIS, s. 1. Honourable birth. Dunbar. 2. Genteel manners. Wallace. 3. Gentleness, softness. Henrysone. GEO, (g hard) s. A deep hollow, Caithn. Isl. gia, hiatus oblongus. 2. A creek or chasm in the shore is called geow, Orkn.
- 43. GER, GERE, GEIR, GEAR, (g hard) s. 1. Warlike accoutrements. Barbour. Isl. geir, lancea; Dan. dyn geira, strepitus armorum. 2. Goods. Goods and gear, a law phrase, S. Ruddiman. 3. Booty. Minst. Bord. 4. All kind of tools for business, S. Ruddiman. 5. Money, S. Watson. Gerit, Geared, part. adj. Provided with armour. Wallace. GERLETROCH. s. V. Gallytrough. GERRON, GAIRUN, s. A sea-trout, Ang. Minst. Bord. GERS, GYRS, s. Grass, S. Wyntown. A. S. gaers, Belg. gars, gers, id. Gersy, adj. Grassy, S. Douglas. Gerss-house, s. A house possessed by a tenant who has no land attached to it, Ang. Gersslouper, s. A grasshoper, S. B.
- 44. Gerss-man, Grass-man, s. A tenant who has no land. Spalding. Su. G. graessaeti, id. Gerss-tack, s. The lease which a gerss-man has, Ang. GERSOME, GRESSOUME, s. A sum paid to a landlord by a tenant, at the entry of a lease, or by a new heir to a lease or feu, S. Dunbar. A. S. gaersuma, gersume, a compensation. To GES, v. n. To guess. Wyntown. GESNING, GESTNING, s (g hard) Hospitable reception. Douglas. Isl. gistning, id. from gest-r, a guest. GESSERANT, Sparkling. K. Quair. Teut. ghester, a spark. GEST, s. Ghost. V. Gaist. Houlate. GET, GETT, GEAT, GEIT, s. 1. A child. Wyntown. 2. A contemptuous designation for a child, S.
- 45. Knox. 3. Progeny. Wyntown. 4. Applied to the young of brutes. Goth. get-a, gignere. Douglas. GEWE, conj. If. V. Gif. To GY, GYE, v. a. To guide. K. Quair. O. Fr. guier, id. Gy, s. A guide. Hisp. guia. Wallace. GY, s. A proper name; Guy, Earl of Warwick. Bannatyne Poems. GIB, GIBBIE, (g hard), s. A gelded cat, S. Fr. gibb-ier, to hunt. Henrysone. GIBBLE, (g hard), s. A tool of any kind, S.; whence giblet, any small iron tool, Ang. Teut. gaffel, furca. Morison.
- 46. GIBBLE-GABBLE, s. Noisy confused talk, S. Isl. gafla, blaterare. Gl. Shirr. GIDE, GYDE, s. Attire. Wallace. A. S. giwaede, id. To GIE, v. a. To give, S. V. Gif. GIELAINGER, s. A cheat. V. Gileynour. GIEST, A contr. of give us it, S. Henrysone. To GIF, Gyf, Giff, v. a. To give; gie, S. Barbour. GIF, GYVE, GEUE, GEWE, conj. If. Douglas. Moes. G. gau, id. Su. G. jef, dubium. GIFFIS, GYFFIS, imper. v. Gif. Douglas. GIFF-GAFF, s. Mutual giving, S. Kelly.
- 47. A. S. gif and gaf, q. I gave, he gave. GYIS, GYSS, s. 1. A mask. Dunbar. 2. A dance after some particular mode. O. Fr. gise. Henrysone. GYKAT. L. Gillot. Maitland P. GIL, (g hard), s. A cavern. Douglas. Isl. gil, hiatus montium. GILD, s. Clamour, noise. A. Hume. Isl. gelld, clamor; giel, vocifero. Gild, adj. Loud, S. B. GILD, adj. 1. Strong, well-grown. Skene. Su. G. gild, validus, robustus. 2. Great. A gild rogue, a great wag. Ruddiman.
- 48. GILD, GILDE, s. A fraternity instituted for some particular purpose, S. Stat. Gild. A. S. gild, fraternitas, sodalitium. Gild-brother, s. A member of the gild, S. GILDEE, s. The whiting pout. Statist. Acc. GYLE-FAT, s. The vat used for fermenting wort, S. Gyle, Orkn. Burrow Lawes. Teut. ghijl, cremor cerevisiae. GILEYNOUR, GILAINGER, s. 1. A deceiver. Kelly. 2. "An ill debtor." Gl. Ramsay. Su. G. gil-ia, to deceive, gyllningar, fraudes. GILLIE, s. 1. A boy. S. P. Repr. Ir. gilla, giolla, a boy; a servant, a page. 2. A youth who acts as a servant, page, or constant attendant, S. Rob Roy. GILLIEGAPUS, GILLIEGACUS. V. Gapus.
- 49. GILLIEWETFOOT, GILLIEWHIT, (g hard) s. 1. A worthless fellow, who gets into debt and runs off, Loth. 2. A running footman; also, a bum-bailiff. Colvil. From gillie, a page, and wet foot. GILL-WHEEP, GELL-WHEEP, s. 1. A cheat, S. B. Shirrefs. 2. To get the gill-wheep, to be jilted, S. B. Isl. gil-ia, amoribus circumvenire, and hwipp, celer cursus. GYLMIR. V. Gimmer. GILPY, GILPEY, s. A roguish boy, a frolicsome boy or girl, S. Ramsay. A. S. gilp, ostentation, arrogance. GILSE, s. A young salmon. V. Grilse. GILT, pret. v. Been guilty. K. Quair. A. S. gylt-an, reum facere. GILT, s. Money. S. gelt. Watson. Germ. gelt, id. from gelt-en, to pay.
- 50. GILTY, adj. Gilded. Douglas. GYM, adj. Neat, spruce, S. Doug. GIMMER, GYLMYR, (g hard) s. 1. A ewe that is two years old, S. Compl. S. Su. G. gimmer, ovicula, quae semel peperit. 2. A contemptuous term for a woman, S. Ferguson. GYMMER, compar. of Gym. Evergreen. To GYMP, (g soft) v. n. To gibe, to taunt. Ruddiman. Isl. skimp-a, Su. G. skymf-a, to taunt. Gymp, Jymp, s. 1. A witty jest, a taunt, S. B. Douglas. 2. A quirk, a subtilty. Henrysone. Belg. schimp, a jest, a cavil. GYMP, GIMP, JIMP, adj. 1. Slim, delicate, S. Douglas. 2. Short, scanty, S.
- 51. Su. G. skamt, short, skaemt-a, to shorten. Gimply, Jimply, adv. Scarcely, S. GIN, conj. If, S. Sel. Ball. GYN, GENE, s. 1. Engine for war. Barbour. Gynnys for crakys, great guns. Barbour. 2. The bolt or lock of a door, S. Ruddiman. GYN, s. A chasm. Douglas. A. S. gin, hiatus. To GYN, v. n. To begin. K. Quair. Gynnyng, s. Beginning. Wyntown. GINGE-BRED, s. Gingerbread, S. Pitscottie. GINKER, s. A dancer. Watson. Germ. schwinck-en, celeriter movere.
- 52. GYNKIE, (g hard) s. A term of reproach applied to a woman; a giglet, Renfr. Ang. Isl. ginn-a, decipere. GYNOUR, s. Engineer. Barbour. GIPE, s. One who is greedy or avaritious. Isl. gypa, vorax. Watson. GIPSY, s. A woman's cap, S. Gipsey herring, The pilchard, S. Ess. Highl. Soc. GIRD, GYRD, s. 1. A hoop, S.; also girr. Minst. Bord. A. S. gyrd, Isl. girde, vimen. Girder, s. A cooper, Loth. 2. A stroke, S. Barbour. To let gird, 1. To strike. Chr. Kirk. 2. To let fly. Douglas. To Gird, v. a. 1. To strike, with the pron. throw.
- 53. Douglas. To Gird, v. n. To move with expedition and force. Barbour. To GIRD, v. n. To drink hard, S. B. Forbes. GIRD, s. A trick. Douglas. Su. G. goer-a, incantare; utgiord, magical art. GIRDLE, s. A circular plate of malleable or cast iron, for toasting cakes over the fire, S. Colvil. Su. G. grissel, the shovel used for the oven; from graedd-a, to bake. GYRE-CARLING, (g hard) s. 1. Hecate, or the mother-witch of the peasants, S. Lyndsay. Gy-carlin, Fife.; Gay-carlin, Bord. Isl. Geira, the name of one of the Fates, and karlinna, an old woman. 2. A hobgoblin. Bannat. Journal. 3. A scarecrow, S. B. Journal Lond. GYRE FALCON, s. A large hawk. Houlate.
- 54. Germ. geir, a vulture, and falke, a falcon. GYRIE, (g soft) s. A stratagem, Selkirks. V. Ingyre. To GIRG, JIRK, v. n. To make a creaking noise, S. V. Chirk. Douglas. GIRKE, s. A stroke, E. jerk. Z. Boyd. Isl. jarke, pes feriens. To GIRN, v. n. 1. To grin, S. Douglas. 2. To snarl, S. Ramsay. 3. To gape; applied to dress, S. Girn, s. A grin, S. Gyrning, s. Grinning. Barbour. GIRN, GYRNE, s. 1. A grin, S. Bellenden 2. A snare of any kind. Ramsay. A. S. girn, Isl. girne, id.
- 55. GIRN, s. A tent put into a wound, a seton, Bord. Isl. girne, chorda. GIRNALL, GIRNELL, GRAINEL, s. 1. A granary, S. Knox. Girnal-ryver, the robber of a granary. Evergreen. 2. A large chest for holding meal, S. Fr. grenier, id. To Girnal, v. a. To store up in granaries, S. Acts Ja. II. GIRNIGO, GIRNIGAE, s. A contemptuous term for a peevish person, S. Gl. Complaynt. GIRNOT, s. The gray Gurnard; vulgarly garnet, Loth. Statist. Acc. GYRS, s. Grass. V. Gers. GIRSILL, s. A salmon not fully grown. Acts Ja. III. GIRSLE, s. Gristle, S. Girslie, adj. Gristly, S. J. Nicol.
- 56. GIRT, pret. v. Made, for gert. Houlate. GIRTEN, s. A garter. Burel. GIRTH, GYRTH, GIRTHOL, s. 1. Protection. Wallace. 2. A sanctuary. Barbour. 3. The privilege granted to criminals during certain holidays. Baron Court. 4. Metaph. in the sense of privilege. Wyntown. To GYS, v. a. To disguise. V. Gyis. GYSAR, GYSARD, s. 1. A harlequin; a term applied to those who disguise themselves about the time of the new year, S. gysart. Maitland Poems. 2. One whose looks are disfigured by age, or otherwise, S. Journal Lond. To GYSEN. V. Geize. GISSARME, GISSARNE, GITHERN, s. A hand-ax, a bill. Doug.
- 57. O. Fr. gisarme, hallebard; from Lat. gesa, hasta, Roquefort. GITE, s. A gown. Chauc. id. Henrysone. GYTE. To gang gite, to act extravagantly, S. hite, S. B. Ramsay. Isl. gaet-ast, laetari. GITHERN. V. Gissarme. Douglas. GYTHORN, s. A guitar. Houlate. Fr. giterne, from Lat. cithara. GITIE, adj. Shining as agate. Watson. GIZZEN, s. Childbed. V. Jizzen-bed. To GIZZEN, v. n. To be dried. V. Geyze. To GLABBER, GLEBBER, v. n. To speak indistinctly, S. Gael. glafaire, a babbler.
- 58. GLACK, s. 1. A defile between mountains, Perths. Ang. Minstrelsy Bord. 2. A ravine in a mountain. Pop. Ball. 3. An opening in a wood where the wind comes with force, Perths. 4. The part of a tree where a bough branches out. Gl. Pop. Ball. 5. That part of the hand between the thumb and fingers. Ibid. Gael. glac, a narrow glen, glaic, a defile. GLACK, s. 1. A handful or small portion, Ang. Ross. 2. As much grain as a reaper holds in his hand, Ang. 3. A snatch, a slight repast, Ang. Gael. glaic, a handful. To GLACK one's mitten, to put money into one's hand, S. B. Journal Lond. Gael. glac-am, to receive. GLAD, GLAID, GLADE, GLID, adj. 1. Smooth, easy in motion, S. Ruddiman. 2. Slippery; glid ice, S. B. 3. Applied to one who is not to be trusted, S. B. A. S. glid, Belg. glad, Su. G. glatt, lubricus. GLADDERIT, part. pa. Besmeared. Teut. kladder-en, to bedaub.
- 59. Welcome to our website – the ideal destination for book lovers and knowledge seekers. With a mission to inspire endlessly, we offer a vast collection of books, ranging from classic literary works to specialized publications, self-development books, and children's literature. Each book is a new journey of discovery, expanding knowledge and enriching the soul of the reade Our website is not just a platform for buying books, but a bridge connecting readers to the timeless values of culture and wisdom. With an elegant, user-friendly interface and an intelligent search system, we are committed to providing a quick and convenient shopping experience. Additionally, our special promotions and home delivery services ensure that you save time and fully enjoy the joy of reading. Let us accompany you on the journey of exploring knowledge and personal growth! ebookname.com