Service oriented cloud architecture for improved

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 02 Issue: 09 | Sep-2013, Available @ http://www.ijret.org 1
SERVICE ORIENTED CLOUD ARCHITECTURE FOR IMPROVED
PERFORMANCE OF SMART GRID APPLICATIONS
Rajeev T.1
, Ashok S.2
1
Research Scholar, 2
Professor, Department of Electrical Engineering, National Institute of Technology, Calicut, Kerala,
India, mail2rajeevt@gmail.com, ashoks@nitc.ac.in
Abstract
An effective and flexible computational platform is needed for the data coordination and processing associated with real time
operational and application services in smart grid. A server environment where multiple applications are hosted by a common pool of
virtualized server resources demands an open source structure for ensuring operational flexibility. In this paper, open source
architecture is proposed for real time services which involve data coordination and processing. The architecture enables secure and
reliable exchange of information and transactions with users over the internet to support various services. Prioritizing the
applications based on complexity enhances efficiency of resource allocation in such situations. A priority based scheduling algorithm
is proposed in the work for application level performance management in the structure. Analytical model based on queuing theory is
developed for evaluating the performance of the test bed. The implementation is done using open stack cloud and the test results show
a significant gain of 8% with the algorithm.
Index Terms: Service Oriented Architecture, Smart grid, Mean response time, Open stack, Queuing model
---------------------------------------------------------------------***------------------------------------------------------------------------
1. INTRODUCTION
Smart grid is a complex network involving large number of
energy sources, controlling devices and load centers. Now the
focus is on the development of a dynamic smart grid
management platform for offering various services. An
interconnected smart grid with large number of dispersed
renewable energy sources, its associated measuring and
control functionalities require large data storage. The existing
centralized approach is not effective in such situation with
huge data storage and computational needs. The new
infrastructure to replace the existing one should address the
future data storage and computational needs. An efficient
smooth information exchange for monitoring and control of
widely distributed power sources is also needed. The immense
potential of cloud computing technology can be utilized to
address these issues. The sharing of resources in various
substations reduces the cost of operation, improves the
performance of utility and offers environmental friendly smart
grids. The cloud environment provides a flexible way of
building, facilitating computing and storage infrastructures for
varying on line and offline services. A flexible and upgradable
cloud computing architecture for application deployment
offers efficient application sharing over the internet.
Modern power system is structured with distributed energy
resources which are required to deal with large amount of data
and information systems [1]-[2]. The storage and processor
resources become increasingly higher with the integration of
renewable sources to the existing grid [3]. Cloud computing in
large power grid and cloud data service center are considered
as one of the central options which can integrate current
infrastructure resources of the enterprise like hardware, high-
performance distributed computing and data platform. The
recent work [4] presented a cloud computing model for
managing the real time streams of smart grid data for the near
real time information retrieval needs of the different energy
market actors. The approaches in [4]-[5] considered the model
of ubiquitous data storage and data access of the smart grid
data cloud, focusing on the characteristics of the underlying
cloud computing techniques. Architecture for data storage,
resource allocation and power management and control is
presented in [6]. The paper discusses existing issues and
necessity of a cloud computing architecture for power
management of micro grids.
Efficient utilization of resources is important in cloud
computing and for that scheduling plays a vital role to get
maximum benefit from resources [7]-[10]. Wei-Tek Tsai et al.
(2010) illustrated the Service-Oriented cloud computing
architecture. Cloud computing is getting popular and IT giants
such as Google, Amazon, Microsoft, IBM have started their
cloud computing infrastructure. The paper gives an overview
survey of current cloud computing architectures and discusses
the existing issues of current cloud computing
implementation. They presented a Service-Oriented Cloud
Computing Architecture (SOCCA), so that clouds can
interoperate with each other. Furthermore, the SOCCA also
proposes high level designs to support multi-tenancy feature
of cloud computing.

__________________________________________________________________________________________
In line with the above, the paper presents general service
oriented architecture for real time services in a distributed
structure. The architecture for real time services generally
needs to incorporate various updates in the service level as
well as at the architectural level. Hence an open source
structure is preferred for the implementation set up. A priority
scheduling algorithm is presented in this paper for
performance management. Analytical model based on queuing
theory to capture the performance of the above structure is
also developed for evaluation.
2. SERVICE ORIENTED ARCHITECTURE FOR
REAL TIME SERVICE
Cloud computing is emerging with rapidly growing service
oriented architectures for real time services. Service oriented
architectures for various online services require a dynamic
adaptable infrastructure for sharing data, compute and
transaction services across applications. The system
architecture proposed in the paper enables secure and reliable
movement of information and transactions with the internal
resources, also with users over the internet to support various
services. The middleware provides the necessary functionality
required to build and deploy fully operational application
services. It includes resource management, data management
and portability. The logical service model is shown in
Figure1.The infrastructure as a service provides virtualized
data storage and computing power. The smart grid
applications in the cloud involves on line and offline
computations, monitoring and analysis of large data, including
online measurement data. The proposed architecture takes full
advantage of functionality provided by open stack and ensures
real time operations in a widely distributed environment.
Operating system requires hypervisor for creation and
termination of instances.
Fig -1: Logical View of Cloud Test bed for Smart grid
The process for selecting a hypervisor means priority and
making decisions based on resource constraints, numerous
supported features and required technical specifications.
Kernel Virtual Machine is selected as hypervisor in the
architecture and there is flexibility for selecting multiple
hypervisors for different zones. Different algorithms and
software associated with execution of the on line pricing and
data analytical operations are deployed as instances in the
smart grid application layer. The service scheduler is realized
through open stack nova, allocates the request according to the
service priority.
Each virtual machine shares resources on a physical server,
including CPU capacity, disk access bandwidth and network
I/O bandwidth. Hence the general terminology resource is
used throughout this paper to represent all the above shared
parameter together.
2.1 Implementation
The details of specifications of the test bed are shown in Table
1. Laboratory test bed has been set up for realizing the open
source cloud computing environment for various data
intensive and computational intensive applications in real time
mode. The performance of the architecture for different test
cases was recorded using web stress tool. The architecture for
meeting the above goal is depicted in Figure 2. The
components of open stack [12] configured for services include
nova-API nova-compute, nova-scheduler, nova-volume and
nova-network. Nova-API initiates most of the activities such
as running an instance and provides an endpoint for all API
queries. Nova-scheduler process take a virtual machine
instance request from the queue and determines where it
should run. The creation and termination of virtual machine
instance is controlled by nova-compute process. It accepts
actions from the queue and then performs a series of system
commands like launching a Kernel Virtual Machine (KVM)
instance to carry out while updating state in the database.
Table- 1: Recommended Hardware/Software
Item Recommended Hardware/Software
Cloud
Controller
HpProliant,64bit x86,2048KCache,12
GB RAM,2x1TBHDD,1GB NIC
Client Node Intel Pentium 4, 3.20 GHz; 2048K
Cache,8GB RAM, 1TB HDD,2X1GB
NIC
Operating
System
Ubuntu12.04
Middleware Open stack
Monitoring Web Stress Tool

__________________________________________________________________________________________
Fig -2: Architectural frame work for Application management
The nova-volume manages the creation, attaching and
detaching of persistent volumes to compute instances. Nova-
network accepts networking tasks from the queue and then
performs tasks to manipulate the network. Data base for
storing generation and consumption profile, the data handling
service for aggregating data from various locations, were
created as instances in the framework.
The architecture proposed here is for the operation of various
application services related to smart grid. The test case
considers application services relating to power system with
distributed generation to simulate the interaction between the
generating sources, loads and consumers. The data related to
various services were stored in the database created as
multiple instances. The aspect of real time data manipulation
was executed through software programs. The data handling
service created as another instance, manages the data traffic.
The case considered for testing involves N consumers and M
number of generating sources. Each production/consumption
update is forwarded to the multiple instances using data
handling service. Though it is an approximation of data
sending and reception by smart meters in the utility and
consumers, it contains all the required components and price
dynamics that are likely to be present in the future smart grid.
Real time pricing and data analytic services were deployed in
different virtual machines. Pricing calculation service involve
several algorithmic steps to reach its decision where as simple
data retrieval and processings are involved in the other
category. To avoid traffic conjunction, the applications which
require huge computing will be given lower priority. In the
architecture, all the request which is coming from client nodes
are goes through the service scheduler. It is preloaded with the
priority status of various services. Here the scheduler simply
allocates to the lightly loaded virtual CPU during normal
situations and executes priority inversion, when admissible
delay crosses the point. Python script has been used to
implement the algorithm.
The timely data retrieval and computational requirements
associated with the execution of on line enquiry from user
perspective was tested for different values of request rate. It
has been found that the response time requirement for real
user interactive services is very demanding. Moreover, in
virtualized environments resource allocation actuation can be
done inside the server as well, using interfaces available from
virtual machine monitors or hypervisors. Response time and
CPU consumption are used are reference parameters in such
operations. Hence in each experiment, time series of response
times and CPU consumption were collected and analyzed. The
performance results for 50,000 consumers requesting for
services at a request rate of 30/s is shown in Figure 3and
Figure 4, where actual CPU resources consumed by virtual
machines are represented as CPU consumption. The rate of
request considered here is comparable with the average
request level for above category of services in smart grid
environment with such number of consumers. The mean
response time varied between 35ms-45ms. The average value
of CPU consumption was about 50%-60%. These values are
reasonable considering the smaller test bed and the complexity
of applications used in the test. To have a better understanding
of the characteristics of the system for massive hit rate and
higher utilization of resources, another set of tests were
conducted by simulating cases with 100% incease in the
resource demand with same resource level. Figure 5 shows
the performance change in the test bed for higher utilization of
resources. Here, the percentage change Mean Response Time
(MRT) is defined as the percentage decrease in response time
for a given resource level and 100% increase in demand. High
utilization of the available resources while keeping the good
response is one of the advantages of the proposed algorithm.
The consistency in the performance was tested by conducting
experiments in the test bed with different cases. Table 2
depicts percentage of failed hits for different degree of
concurrent request. Tests were conducted by applying linearly
increasing request level for various numbers of users
considered. It has been found that the algorithmic approach
maintains good success rate for varying level of request rate.
Table- 2: Statistics of failed hits
Number
of users
%Error(Failed Hits)
With Priority
Algorithm
Without Algorithm
1000 0.5 0.6
5000 1.21 2.1
10000 5.1 7.2
15000 7.2 11.3
20000 7.9 12.1
25000 8.4 19.5
30000 11.5 21.8

__________________________________________________________________________________________
35000 12.5 23.7
40000 14.4 25.8
45000 15.7 26.9
50000 17.5 31.8
Fig -3: Mean response Time
Fig -4: Time varying demand of CPU
Fig -5: Performance change with same resource level
3. DISCUSSION
The objective of these experiments was to test the
effectiveness of the architecture to handle real time services,
in smart grid environment. The study focuses on the
performance metrics. It is apparent from various results that
the proposed architecture in open stack cloud environment
maintains satisfactory level of responses for varying request
rate. Also found that performance is affected by the resource
capacity available to the system and the time varying demand
of the applications. The cloud environment succeeds to limit
the failure rate and offered a reasonable value of MRT for the
cases considered. Scheduling the applications according to
priority aids the dynamic resource allocation in virtualized
server environment. Result reveals that, at higher values of
resource demand, resource utilization is dynamically adjusted
and resources are allocated on time. The open source software
considered in the experiment permits up gradation of existing
architecture for incorporating more features. The architecture
can be extended to incorporate scheduling algorithms based on
other performance metrics. The analytical model developed
for evaluating the performance of the architecture is described
in the next section. The model considers the characteristics of
resource allocation in a dynamically changing workload
situation
4. ANALYTICAL MODEL FOR PERFORMANCE
EVALUATION.
The analytical modeling proposed in the work capture the
performance of the multiple applications hosted in a
virtualized environment. The work considers queuing theory
based modeling to characterize the performance of cloud
environment, where applications are distributed across
multiple virtual machines with time varying resource demand.
The modeling also uses the terminology resource to represent
all the resources shared by a virtual machine on a physical
server, e.g., resource consumption is considered in place of
CPU consumption. The paper [13] presented an analytical
model for multi-tier Internet application based on a network of
queues to represent how the tiers in multitier application
cooperate to process requests and demonstrated the utility of
the model in managing resources for internet application under
varying workloads and shifting bottlenecks. However, a
virtual machine differs from a physical server in that its
effective capacity varies with dynamic resource allocation. In
dynamically changing workload situations, the optimal
performance can be achieved by allotting the application
request based on some priority. The approach used in this
work is to categorize the different applications deployed in the
deployment platform as high priority and low priority ones.
The operational mode in the algorithm involves customer
arrival ruled by the distribution of inter arrival time according
to a poison process. The aggregate user request rate, R, is
defined by equation (1), where N is the number of different

__________________________________________________________________________________________
applications deployed and Rn is the mean user request rate of
application n.
∑=
=
N
1n
nRR
The queue discipline is according to a priority with negligible
delay and the resources in the virtualized environment can be
modeled as M/G/1/PRIOR queue [14]. According to queuing
theory, total resource resident time by all the request served in
tier k is represented by qk/(1-qk), where qk is the resource
utilization in tier k. For an aggregate user request rate of R, the
mean resident time for applications with higher priority, Tsh
can be described as






−
=
k
k
sh
q1
q
R
1
T
Resource consumption for a typical application is proportional
to number of request. Let Dn represent the mean demand of
applications with higher priorities. Then resource consumption
by highest priority application, Mc, can be defined as a linear
function of user request rate.
∑=
=
N
1n
nnkc R*DM
The resource utilization by all the application in the virtualized
environment is defined as the ratio between the virtual
machine’s resource consumption and its effective resource
capacity. In the virtualized environment resource capacity is
dynamically modified. The utilization changes according to
the changes in allocation. Hence resource utilization is
modified as the ratio between resource consumption of the
virtual machine and resource allocation to the virtual machine.
Let Ma be the resource allocation, refers to the resource
capacity that is allocated to a virtual machine. Incorporating
the modification in (2),Tsh can be represented as
∑
∑
∑
=
=
=
−
= N
1n
nnka
N
1n
nnkK
1k
sh
R*DM
R*D
R
1
T
The modeling assumes applications with low priority are
available with maximum resources. Since priority inversion is
considered in the algorithm the modeling considers total
number of applications for the calculation resident time in this
case also. Mean resident time for applications with low
priority is represented by Tsl, can be approximated as (5),
where values of φn represent mean resident time for
applications with lower priorities. The value can be obtained
through model calibration. Combining (4) and (5) the mean
response time can be represented by aggregating resident
times over all resourses by higher priority and lower priority
applications. Thus once the values of average demand and
mean resident time for applications with lower priorities are
obtained, mean response time can be predicted from equation
(6).
∑=
=
N
1n
nnsl R*φ
R
1
T
slsh TTMRT +=
The linear regression method is used to estimate the mean
demand from (3) and response time from equation (6) for a
given arrival rate of request. The estimate considered average
user request rate of 1800- 2400 per minute and a mean
response time target of 35ms to 45ms.
The measured values of MRT were compared with those
estimated using (6) for evaluation. The performance result in
the test for normal and high utilization of resources is depicted
along with estimated values in figure 6 and figure 7.The model
is valid for the present case as it captures the effect of resource
capacity, time varying demand and response with respect to
complexity in applications. The estimate considered the same
entitlement and user request rate as used in the test cases. The
error between the measured and estimated values of response
times for various cases is negligible. The result justified the
architectural performance.
The modeling used in the work is application oriented. The
model can predict the performance of the cloud environment
with various applications, running on the virtualized servers
having different computational needs. The modeling
considered here needs to include more parameters for a
complex networked structure. In such cases, advanced non
linear solution methods provide precise results than regression
method.
Fig -6: Response under normal utilization
(1)
(2)
(3)
(4)
(5)
(6)

__________________________________________________________________________________________
Fig -7: Response under high utilization
CONCLUSIONS
The architecture explores a simplified approach through
functional building blocks; all aimed at delivering real time
services in service oriented structure. It support creation,
execution and evolution of service oriented solutions. A
priority scheduling algorithm was applied to the architecture
for performance management.
The proposed algorithm was tested with the application
services relating to power grid which has time varying
processing and storage need. The result showed that the
architecture have better performance with the algorithm.
Analytical model based on queuing theory to capture the
performance of the architecture has been developed for
evaluation. The result justified that the proposed architecture
is effective for real time applications services in smart grid
environment.
The work is continuing with the modifications in the
architecture and algorithm to develop load management
system for smart grid that can optimally coordinate timely
demand side management.
REFERENCES
[1] G.T.Heydt, B.H.Chowdhury, M.L.Crow,
D.Haughton,B.D.Kiefer,F. Meng and
B.R.Sathyanarayana “Pricing and Control in the Next
generation Power Distribution System” IEEE
Transactions on Smart Grid,Vol. 3,No. 2,June
2012,pp.907-914
[2] F.Katiraei and M.R.Iravani, “Power Management
Strategies for a Microgrid With Multiple Distributed
Generation Units”, IEEE Transactions on Power
Systems,Vol.22,No.4,November 2006,pp.1821-1831
[3] Soma Shekara Sreenadh reddy, Depura, Lingfeng
wang, vijay Devabhaktuni, “Smart meters for Power
grid Challenges,issues,advantages and Status”,
Renewable and Sustainable energy Reviews,
Vol.15,2011,pp.2736-2742.
[4] R Rusitschka, K. Eger, C. Gerdes, "Smart Grid Data
Cloud: A Model for Utilizing Cloud Computing in
the Smart Grid Domain", in Proceedings IEEE
International conference on Smart Grid
Communications,2010., pp.-483-488
[5] Amir-Hamed Mohsenian-Rad, Albert Leo-Garcia ,
“Cordination of Cloud Computing and Smart Power
grids”, in: Proceedings IEEE International conference
on Smart grid communications, 2010, pp. 368-372.
[6] Rajeev T., Ashok S, A Cloud Computing Approach
for Power Management of Micro grids, in:
Proceedings IEEE PES Innovative Smart grid
Technologies, India (ISGT-India) 2011, pp. 49-52
[7] Pinal Salot, “A survey of various scheduling
algorithms in Cloud Computing Environment”,
International Journal of Research in Engineering&
Technology( IJRET), Vol. 2,No. 2,February
2013,pp.131-135
[8] Zhikui Wang, Yuan Chen, Daniel Gmach, Sharad
Singhal, Brian J. Watson,Wilson rivera,Xiaoyun Zhu,
and Chris D. hyser, “AppRAISE:Application-Level
Performance Management in Virtualized Server
Environments,” IEEE Transactions on Network and
service management ,Vol. 6, No.4.Dec 2009.pp.240-
253.
[9] Ruben Van den Bossche , Kurt Vanmechelen, Jan
Broeckhove, “Online cost-efficient scheduling of
deadline-constrained workloads on hybrid clouds”,
Future Generation Computer Systems, Vol.29, May
2013,pp. 973-985
[10] Brijesh Goyal,Pallavi Jain, “Reminiscing Cloud
Computing Technology”, International Journal of
Research in Engineering& Technology( IJRET),Vol.
1,No. 2,November 2012,pp. 364-367
[11] Wei-Tek Tsai, Xin Sun, Janaka Balasooriya, “
Service-Oriented Cloud Computing
Architecture”in:,Proceedings on Seventh
International Conference on Information
Technology,April,2010.pp. 684-689.
[12] OpenStack Beginner's Guide(for Ubuntu - Precise)
v3.0, 7 May 2012
[13] B.Urgaonkar,G.Pacifici,P.Shenoy,M.Spreitzer,and
A.Tantawi, “An analytical model for multi-tier
internet services and its applications,”ACM

__________________________________________________________________________________________
SIGMETRICS Performance Evaluation Review,. Vol
33,June 2005,pp.291-302.
[14] E.D.Lazowska, J.Zahorjan, G.S.Graham, and
K.C.Sevcik, uantitative Sysyem Performance:
Computer System Analysis Using Queuing Network
Models.upper Saddle River,N.J:Prentice-
Hall,Inc.,1984.

Service oriented cloud architecture for improved

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Service oriented cloud architecture for improved (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Service oriented cloud architecture for improved