Fuzzy Based Algorithm for Cloud Resource Management and Task Scheduling

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 3 Issue 5, August 2019 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD26604 | Volume – 3 | Issue – 5 | July - August 2019 Page 1335
Fuzzy Based Algorithm for
Cloud Resource Management and Task Scheduling
Rashmi Singh Lodhi1, Dr. R. K. Pateriya2
1PhD Scholar, 2Associate Professor
1,2Manit, Bhopal, Madhya Pradesh, India
How to cite this paper: Rashmi Singh
Lodhi | Dr. R. K. Pateriya "Fuzzy Based
Algorithm for Cloud Resource
Management and Task Scheduling"
Published in International Journal of
Trend in Scientific
Research and
Development
(ijtsrd), ISSN: 2456-
6470, Volume-3 |
Issue-5, August
2019, pp.1335-
1340,
https://doi.org/10.31142/ijtsrd26604
Copyright © 2019 by author(s) and
International Journal ofTrend inScientific
Research and Development Journal. This
is an Open Access article distributed
under the terms of
the Creative
Commons Attribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by
/4.0)
ABSTRACT
This paper presents a Fuzzy Logic based approach to manage VM status and
VM configuration within the cloud environment. The aim of the approach is to
serve task requests efficiently with minimal use of resource and power. The
proposed technique uses Fuzzy based approach to calculate the VM’s status
and VM's configurations, depending upon the cloud resource availability and
job resource requirements, and then a fuzzy logic-based controller is used to
control the status and configurations of the VM to serve the intended purpose
afterwards. Controlling in this way reduces the active physical resources and
the cloud's power requirements.The proposed controller is testedfordifferent
load conditions against the standard controlling algorithm to validate the
concept. The test results obtained show that in terms of QoS, resource
management, and power savings, the proposed fuzzy logic controller-based
technique outperforms standard techniques.
KEYWORDS: Cloud Computing, Cloud Resource Management, Cloud VM
Management, Fuzzy Logic Controller
1. INTRODUCTION
In a cloud system the Virtual Machines (VMs) that are formed within the cloud
system, are used to run the user services or applications. The VMs are formed
by virtually allocating and configuring thephysical resources.Thevirtualization
platforms are used to manage these VMs. The cloud service providers need to
maintain a minimum level of Quality of Service (QoS) bound by the Service
Level Agreement (SLA) which is also known as the Service Level Objective
(SLO).
Increasing application of cloud systems, imposing new
challenges for cloud operators as they now have todeal with
limited resources with greater loads of differentoperational
and requirement features. To tackle such situations, an
efficient cloud management algorithm is needed that could
utilized cloud resources to meet the job requirements with
guaranteed SLO. while keeping power requirements at the
lowest possible level.
This paper presents a Fuzzy Logic based VM status and
configuration management technique. The proposed
technique uses task and cloudcharacteristics,thenusingthis
information to manage the VM's status and configuration
through a fuzzy decision-making system based on logic. The
proposed technique can manage the balance between
conditions such as task requirements, SLO boundaries,
sources available, power savings, etc. by dynamically
configuring the VMs.
The rest of the paper is organized as follows. Section II
provides a brief literature review. Section III provides an
overview of Fuzzy's logic controller. Section IV explains the
system architecture withsimulationconfigurations,explains
the proposed system in Section V. Section VI presents the
results of the simulation and finally Section VII presents the
conclusion and discusses the possibilities of future work.
2. LITERATURE REVIEW
The majority of proposed frameworks utilize resource
controllers that are centralized in the sense that they have
full control of the allocation of a given set of resources. A
small number of frameworks incorporate decentralized
controllers. Global scheduling involves a system-wide
perspective on the allocation of the physical and virtualized
resources that comprise a cloud environment. Recent years
have seen a proliferation of proposals for frameworks and
techniques for global scheduling. Most proposals originate
from academic researchers; notable exceptions include the
discussion by Gulati et al. [1] of the VMware Distributed
Resource Scheduler and Distributed Power Management
systems, which provide centralized control of a cluster of
virtualized servers, performing VM placement, as well as
load balancing and VM consolidation via live migration.
Similarly, Wilkes [2] provides a high-level overview of two
generations of Google’s cluster management systems,
focusing on design goals, management objectives and
unsolved challenges.
Frameworks from the academiccommunitythatincorporate
controllers directly or indirectly cooperating to achieve
global scheduling include: vManage [3], which loosely
couple’s controllers forpowerandcoolingmanagementwith
controllers for virtualization management; the system
proposed by Beloglazov and Buyya [4], which comprises
IJTSRD26604

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
controllers that cooperativelycontrolVMplacementandlive
migration; and TROPIC [5], which performs transactional
cloud resource orchestrations that enforce safety, provide
concurrency and increase system robustness. Whenever a
request for the provisioning of one ormoreVMs ismadebya
Cloud User, the Cloud Provider’s resource management
system schedules the VMs by placing them onto PMs—the
decision should help achieve the Cloud Provider’s current
management objective(s). Initial work on the VM placement
problem typically assumed that VMs are assigned static
shares of a PM’s managed resources (typically CPU and/or
memory). Given this, the placement of VMs onto PMs is
related to the vector bin packing problem,whichcanbeused
to model static resource allocation problems where the
resources used by each item are additive. To achieve server
consolidation, for instance, the optimal placement is one
where the items (i.e., VMs) are packed into a minimum
number of bins (i.e., PMs), such that the vector sum of the
items received by any bin does not exceed the bin’s
(resource) limit. The vector bin packing problem and its
variants are NP-hard problems [6]; thus, heuristic
algorithms are commonly proposed and evaluated. Most
proposed heuristics are based on greedy algorithms using
simple rules, such as First Fit Decreasing (FFD) and Best Fit
Decreasing. Panigrahy et al. [7] study variants of the FFD
algorithm and propose a new geometric heuristic algorithm
which scales to large data centers without a significant
decrease in performance. Wilcox et al. [8] propose a new
generic algorithm, Reordering Grouping Genetic Algorithm
(RGGA), which they apply to VM placement. Other works
proposing bin packing heuristics for placement of VMs on
PMs include Jung et al. [9], Gupta et al. [10], and Li et al.
[140].
3. CLOUD COMPUTING
Cloud computing is a general term for the provision of
hosted services over the Internet. These services are widely
divided into three categories: Infrastructure as a Service
(IaaS), Platform as a Service (PaaS) andSoftwareas a Service
(SaaS). The name cloud computing was inspired bythecloud
symbol which is often used in flowcharts and diagrams to
represent the Internet.
A cloud service has three distinct features that distinguish it
from traditional web hosting. It is sold on demand, typically
by minute or hour; it is elastic— a user can have as much or
as little service as they want at any given time; and the
service is fully managed by the provider (the consumer
needs nothing but a personal computer and internetaccess).
Increasing interest in cloud computing has been accelerated
by significant innovations in virtualization and distributed
computing as well as improved access to high-speed
Internet.
3.1 Deployment Models
There are four different deployment models of cloud
computing:
Public Cloud: Traditional, public or external cloud
computing provides dynamic resources through aninternet,
VPN, or off-site third-party provider that accounts fine-
grained. Public cloud: the external cloud or the public cloud.
Community Cloud: To benefit from cloud computing, if
several organizations share the infrastructure, they can set
up a community cloud. This is a more expensive option
compared to the public cloud, as costs are spread overfewer
users than public clouds. However, this option may provide
increased privacy, security and/or policy compliance.
Hybrid clouds: Hybrid cloud means either two clouds
(public, private, internal or external) or a combination of
virtualized cloud server instances with real physical
hardware. Using physical hardware and virtualized cloud
server instances together, the most appropriate definition
for "Hybrid Cloud" is likely to have one common service.
Two clouds that were joined together call the "combined
cloud."
Private Clouds: Private clouds describe private networking
cloud services. It is made up of apps or virtual machines in
the company's set of hosts. In the area of utility computing,
they offer the advantages of shared hardware costs, the
ability to restore fault and increase and decrease depending
on demand.
4. FUZZY LOGIC SYSTEM
The term "fuzzy" refers to things that are either unclear or
vague. Many times, in the real world we find a situation
when we cannot determine whether the stateis trueorfalse,
their fuzzy logic gives reasoning a veryvaluableflexibility.In
this way, we can consider any situation's inaccuracies and
uncertainties.
1 is the absolute value of truth in the Booleansystemand 0is
the absolute false value. But there is no logic to absolute
truth and absolute false value in the fuzzy system. But there
is also an intermediate value in fuzzy logic that is partially
true and partially false.
4.1 Fuzzy Architecture
Its Architecture contains four parts:
Rule Base: It contains the list of rules and the IF-THEN
conditions on the basis of linguistic informationprovided by
the experts to govern the decision-making system. Recent
developments in fuzzytheoryofferseveraleffectivemethods
for fuzzy controller design and tuning. Most of these
developments reduce the number of rules.
Fuzzification: It is used to convert inputs to fuzzy sets, i.e.
crisp numbers. Basically, crisp inputs are the exact inputs
measured by sensors and transferred to the processing
control system, such as temperature, pressure, rpm, etc.
Inference Engine: It determines the corresponding degree
of the current fuzzy input relative to each rule and
determines which rules are to be fired according to the field
of input. Next, to form the control actions, the fired rules are
combined.
Defuzzification: It is used to convert thefuzzysets obtained
by the inference engine to a crisp value. Several
defuzzification methods areavailableand thebest suited one
is used to reduce the error with a specific expert system.

Figure 1: Architecture of Fuzzy Logic Controller (FLC)
system.
Membership Function
A graph that defines how to map each point in the input
space to the membership value between0 and1.Input space
is often referred to as the universe of discourse or universal
set (u), which contains all possible elements of concern in
each particular application.
Figure 2: Six types of fuzzy membership functions:
triangular (A), z-shape (B), trapezoidal (C), s-shape
(D), sigmoid (E) and Gaussian (F), (MATLAB user’s
guide, 2007).
5. PROPOSED ALGORITHM
5.1. System Architecture
The proposed system architecture is presented in Figure 3.
The system contains four Fuzzy Logic Decision Blocks, two
VM Controlling Blocks and five Information Extraction
Blocks. The working details of each block are as follows:
Task Analyzer: This block is used to extract useful
information from task queue.
VM Analyzer: This block is used toextractusefulinformation
from VMs.
 Task Length: the length of current task in MIPS.
 Task Priority: Execution priority of current task.
 VM Utilization: how many times the particular VM has
accessed during predefined time interval.
 VM Configuration: the resources used by VM.
 VM Load: current load on VM.
Fuzzy Logic Estimator and Controller Blocks: These blocks
are used to make specific decisions based on the inputs
provided using fuzzy logic.
 Fuzzy Task Weight Score Estimator: This block
estimates the task requirementsbasedonthelengthand
priority of the task.
 Fuzzy VM Capability Estimator: This blockestimates the
ability of VM to handle basic VM configuration and load
tasks.
 Fuzzy Task-VM Compatibility Estimator: This block
estimates fitness between VM capabilities and task
requirements.
 Fuzzy VM Status Controller: This block is used to
determine the operational status of VM based on the
resource used by VM and the VM access rate.
Figure 3: Proposed System Architecture

VM Selector: This block is used to select the best VM for the
current task; it takes the Fuzzy Relative Score Estimator
input for all VMs and selects the VM with the highest relative
score.
Fuzzy VM status Controller: it keeps the states of all VM on
the basis of VM status score (𝑉𝑀 ). This blocks uses two
different thresholds 𝑡 and 𝑡 where 0 < 𝑡 <
𝑡 < 1 which are compared against VM status score to
decide the VM status as follows:
𝑉𝑀
𝐾𝑒𝑒𝑝 𝑅𝑢𝑛𝑛𝑖𝑛𝑔, 𝑖𝑓 𝑉𝑀 < 𝑡
𝑆𝑙𝑒𝑒𝑝, 𝑒𝑙𝑠𝑒𝑖𝑓 𝑡 ≥ 𝑉𝑀 > 𝑡
𝐷𝑖𝑠𝑠𝑜𝑙𝑣𝑒, 𝑒𝑙𝑠𝑒 𝑉𝑀 ≥ 𝑡

SIMULATION RESULTS
Following consideration is taken into account when the
numerical models are developed.
 It is assumed that the load balancer knows the
configuration (such as processing capacity, memory,
etc.) of each virtual machine (VM) in the cloud.
 With zero-time delay, the load balancer can get the
operational status of each VM.
 The load balancer takes no time to select and assign the
tasks to VM's.
 The load balancer selects the VM based on the selected
algorithm for the input tasks.
 Each VM has zero boot time, so start executing the
assigned task immediately.
 The incoming task size is considered in MI (million
instructions) units.
 The capacities of the VM are also considered in MIPS
units (million instructions per second).
The evaluation of the proposed algorithm is carried out
using numerical computing software from MATLAB. During
the simulation, the tasks arrive at a rate of π as a Poisson
process. The random length tasks are generated using a
uniform discrete distribution within the provided minimum
and maximum task length limits. The similar way is used to
generate task priorities and to define the execution
capabilities of VM.
5.2. Definition of Evaluation Terms
The following measures are used to evaluate the
performance of the algorithm.
SLO Failure: is defined as a cloud failure to serve the task
within a given time limit (inverse of priority).
SLO Failure Task Length: defines the duration of the SLA
Failure task.
VM Reboots: is bootingVMs fromsleepmode,thisoperation
is required when the already running VMs cannot serve the
current task.
VM Reforms: is the formationofnewVMforms theavailable
unused resources when the current task cannot be handled
by the already formed (running or sleeping) VMs.
Resource Utilization Efficiency: shows how efficiently
cloud resources are used to serve tasks and is calculated as
follows:
𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 𝑈𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐 =
∑ 𝑇𝐿
∑ ∑ 𝐶
× 100
Where
𝑇𝐿 : is the load in cloud at time 𝑖.
𝐴 : is the number of VMs active and running at time 𝑖.
𝐶 : Execution capacity of the 𝑗 VM.
𝑁: is the total simulation time (discrete events of task
arrival).
5.3. Simulation Configurations
To properly simulate the algorithm, some important
parameters are required to configure these parameters and
their values are listed in table 2.
Table2: The simulation parameters and their values.
Configuration Parameter Parameter Value
Total Execution Capacity
Available
100 MIPS
Sampling Window Length 10
Minimum Task Length 100 MI
Minimum Task Execution Time 1 Seconds
Maximum Task Execution Time 10 Seconds
Threshold Sleep 0.5
Threshold Dissolve 0.7
Total Simulation Time 100 Seconds
Simulation Results
The simulation results are presented in graphicalforms.The
results of the proposed algorithm are also compared with
the two standard task scheduling algorithms called Round
Robin and Random Selection.
Figure 4: Plot for number of tasks failed to receive the
requested SLA by cloud due shortage of resources with
respect to simulation time.
Figure 5: Plot for total length of the tasks which failed
to receive the requested SLA by cloud due shortage of
resources with respect to simulation time.

Figure 6: Plot for number of times the VM is rebooted
from the sleep mode for assignment of task with
respect to simulation time.
Figure 7: Plot for number of times the VM is reformed
from the available resources for assignment of task
with respect to simulation time.
Figure 8: Plot showing the variation of the cloud
resource utilization efficiency with respect to
maximum task length.
6. CONCLUSION
We presented the fuzzy logic-based cloud systems task
scheduling and resourcemanagementschemesinthis paper.
The simulations show that the proposed algorithm reduces
the number of tasks failed to achieveSLOandthetotal length
of the task failed to achieve SLO compared to standard
algorithms by a factor of 3.0 and 8.0 (Figure 4 and 5)
respectively. This shows thatthealgorithmrejects taskswith
the lowest length when SLO is not delivered guaranteed.
A number of VM reboots and reforms reduced by a factor of
5.0 and 2.5 (figure. 6 and 7) in the proposed algorithm.
Finally, the efficiency comparison (fig. 8) shows that the
proposed algorithm provides maximum efficiency. It also
shows that the efficiency falls much slowly than the other
algorithm, so it can be said that the algorithm provides a lot
of uniform and stable performance for a wide range of
loading conditions. These results show that the proposed
algorithm adequately handles a cloud system for task
schedulingand resourcemanagementwithlimitedresources
and SLO boundaries.
References
[1] Gulati, A., Holler, A., Ji, M., Shanmuganathan, G.,
Waldspurger, C., Zhu, X.: VMware distributed resource
management: design, implementation, and lessons
learned. VMware Tech. J. 1(1), 45–64 (2012).
[2] John Wilkes, 2011 GAFS Omega.
http://youtu.be/0ZFMlO98Jkc (2011).
[3] Kumar, S., Talwar, V., Kumar, V., Ranganathan, P.,
Schwan, K.: Loosely coupled coordinated management
in virtualized data centers. Clust. Comput. 14(3), 259–
274 (2011).
[4] Beloglazov, A., Buyya, R.: Energy efficient resource
management in virtualized cloud data centers. In:
Proceedings of 10th IEEE/ACM International
Conference on Cluster, Cloud and Grid Computing
(CCGrid 2010), pp. 826–831. IEEE (2010).
[5] Liu, C., Mao, Y., Chen, X., Fernandez, M.F., Loo, B.T., van
der Merwe, J.: TROPIC: transactional resource
orchestration platform in the cloud. In: Proceedings of
2012 USENIX Annual Technical Conference (ATC
2012). USENIX (2012).
[6] Coffman Jr, E.G., Garey, M.R., Johnson, D.S.:
Approximation Algorithms for NP-Hard Problems, pp.
46–93. PWS Publishing Co., Boston (1997).
[7] Panigrahy, R., Talwar, K., Uyeda, L., Wieder, U.:
Heuristics for vector bin packing. Technical report,
Microsoft Research (2011).
[8] Wilcox, D., McNabb, A., Seppi, K.: Solving virtual
machine packing with a reordering grouping genetic
algorithm. In: Proceedings of 2011 IEEE Congress on
Evolutionary Computation (CEC), pp. 362–369. IEEE
(2011).
[9] Jung, G., Joshi, K., Hiltunen, M., Schlichting, R., Pu, C.:
Generating adaptation policies for multitier
applications in consolidated server environments. In:
Proceedings of 2008 International Conference on
Autonomic Computing (ICAC 2008), pp. 23–32. IEEE
(2008).
[10] Gupta, R., Bose, S., Sundarrajan, S., Chebiyam, M.,
Chakrabarti, A.: A two stage heuristic algorithm for
solving the server consolidation problem with item-
item and bin-item incompatibility constraints. In:

Proceedings of 2008 IEEE International Conferenceon
Services Computing (SCC 2008), vol. 2, pp. 39–46.IEEE
(2008).
[11] Li, B., Li, J., Huai, J., Wo, T., Li, Q., Zhong, L.: EnaCloud: an
energy-saving application live placement approachfor
cloud computing environments. In: Proceedings of
2009 IEEE International Conference on Cloud
Computing (CLOUD 2009), pp. 17–24. IEEE (2009).
[12] M. S. Q. Z. Nine, M. A. K. Azad, S. Abdullah, and R. M.
Rahman, “Fuzzy logic based dynamic load balancing in
virtualized data centers,” 2013 IEEE International
Conference on Fuzzy Systems (FUZZ-IEEE), 2013.
[13] A. N. Toosi and R. Buyya, "A Fuzzy Logic-Based
Controller for Cost and EnergyEfficientLoad Balancing
in Geo-distributed Data Centers," 2015 IEEE/ACM 8th
International Conference on Utility and Cloud
Computing (UCC), Limassol, 2015, pp. 186-194.
[14] Javanmardi S., Shojafar M., Amendola D., Cordeschi N.,
Liu H., Abraham A. “Hybrid Job Scheduling Algorithm
for Cloud Computing Environment,” IBICA 2014.
Advances in Intelligent Systems and Computing,
vol303. Springer, Cham.

Fuzzy Based Algorithm for Cloud Resource Management and Task Scheduling

More Related Content

What's hot (16)

Similar to Fuzzy Based Algorithm for Cloud Resource Management and Task Scheduling (20)

More from ijtsrd (20)

Recently uploaded (20)

Fuzzy Based Algorithm for Cloud Resource Management and Task Scheduling