IEEE - 31661

A novel approach for task scheduling in cloud

R. Vijayalakshmi, Mrs. Soma Prathibha

Department of Information Technology
Sri Sairam Engineering College
Chennai, India
[email protected], [email protected]

Abstract—Cloud computing is a popular computing

paradigm which enables delivery of on-demand services over the
internet. A cloud environment consists of multiple customers
requesting for the available resources. Cloud vendors who offer
Infrastructure as a Service should enable efficient management
of the available resources. Proper scheduling in cloud enables
the selection of best suitable resources for task execution. In the
present scenario, allocating the resources efficiently is a
challenging job. Service providers need to ensure that their
resources are utilized properly. A new scheduling algorithm is
proposed to address the challenge of task scheduling in cloud.
The user tasks are prioritized. Based on the priority, the tasks
are assigned to the Virtual Machines. The task with highest
priority is assigned to a Virtual Machine with highest processing
power. The proposed model is simulated using the CloudSim
toolkit. The experimental results reveal that the proposed model
results in lower execution time as the resources are allocated
more efficiently.

Keywords— Task scheduling, Virtual Machines, priority,

CloudSim, execution time.

Fig. 1. Characteristics of cloud

I. INTRODUCTION
In the current scenario, cloud computing[1] has grabbed II. SCHEDULING IN CLOUD
the attention of the IT industry. Cloud is an environment
where a large number of systems are connected through a Scheduling of resources[4][5] is considered very much
network, to provide dynamically scalable infrastructure for essential in cloud computing. In general, task scheduling is
application, data and file storage. Today, more number of the process of mapping the tasks to available resources in
organizations benefit from cloud computing to host their cloud, based on their characteristics. Appropriate scheduling
applications. It eliminates the need to invest in new IT results in efficient utilization of the available resources.
infrastructure[2] that includes network, computers, software Scheduling in cloud can be categorized into the following
and database administrators. It expels the need to create new stages:
Datacenters and manage them, thereby saving a huge amount A. Resource discovering and filtering
of cost. It improves accessibility since any user can access the
The Datacenter Broker discovers all the resources present
services from anywhere. The customers of cloud computing
in the network system and collects the status information
need not have to own the infrastructure and hence need not
related to them.
care about their maintenance. Cloud computing uses a pay-as-
you-use scheme in which resources are provided on-demand B. Resource selection
by cloud vendors[3]. Some cloud vendors include Amazon, The target resource is selected based on certain parameters
Microsoft, IBM and so on. The various services provided by of the task and the resource.
the cloud vendors can be broadly classified as Infrastructure
as a Service, Platform as a Service and Software as a Service. C. Task submission
Some of the characteristics that a cloud vendor must consider The task is submitted to the resource that is selected.
while providing services are shown in Fig.1.

4th ICCCNT 2013

July 4-6, 2013, Tiruchengode, India
 IEEE - 31661

III. CLOUDSIM 2) CloudletScheduler: This class implements different

CloudSim[6] is a simulation toolkit that provides support allocation policies that determine the sharing of processing
for modeling and simulation of virtualized cloud-based power among cloudlets in a virtual machine.
Datacenter environments, including management of memory, 3) Host: This class models a physical resource such as a
virtual machines, bandwidth, storage etc. It has a virtualization computer or storage server. It encapsulates information such
engine that helps in the creation and management of as amount of memory and storage, processing cores, policies
independent, multiple virtualized services. It supports dynamic for allocation of memory and bandwidth to virtual machines.
creation of different kinds of entities. Infrastructure as a
Service(IaaS)[7][8][9] can be simulated by extending the 4) Virtual Machines: This class models a virtual machine.
Datacenter entity of CloudSim. The Datacenter manages large This virtual machine is created on a cloud Host component.
number of Hosts. The Hosts are assigned one or more Virtual 5) VMAllocationpolicy: This class determines the policy
Machines(VMs) based on VMAllocation policy. Cloudlets are for allocating VMs to Hosts based on various criteria. The
allocated to VMs based on scheduling policies defined by the main functionality of this class is to select a Host in the
Cloudlet Scheduler. Datacenter that meets the requirements for deployment of the
A. Features virtual machine.
 CloudSim supports modeling and simulation of large- 6) Datacenter: This class models the core infrastructure
scale cloud computing environment, including level services offered by the cloud service providers. It
Datacenters, Hosts etc. encapsulates a set of Hosts that can either be homogeneous or
 It also supports modeling and simulation of user- heterogeneous.
defined allocation policies. 7) DatacenterBroker: This class is responsible for
 It also supports simulation of network connections mediating between cloud service providers and users. It
using network topology. identifies the suitable service providers and allocates the
Using CloudSim, industry-based developers and researches resources in such a way that the users QoS requirements are
can focus on specific system design issues[10] that they want met.
to investigate, rather than considering the lower level details
about the cloud-based services and infrastructure. C. Modeling the allocation of cloudlet to VM
The key aspect of cloud computing infrastructure is the
deployment of various virtualization technologies and tools.
Cloudlets can be allocated to VMs[11] using the following
steps:

 Cloud users submit Cloudlets and request for VMs.

 The Cloudlet list and the VM request list are passed on
to the Datacenter Broker.
 The DatacenterBroker uses the CloudletScheduler for
scheduling the Cloudlets and the VMAllocationpolicy is
used for allocation of VMs to hosts.
Fig. 3 shows how Cloudlets are allocated to VMs in cloud.

Fig. 2. CloudSim Architecture

The following classes of the CloudSim architecture, shown

in Fig. 2, are mainly used in our proposed work.
B. Major classes used
1) Cloudlet: This class models the cloud-based application
services such as business workflow, social networking and so
on. Cloudlets may also be referred to as tasks.
Fig. 3. Allocation of Cloudlets to VM

4th ICCCNT 2013

July 4-6, 2013, Tiruchengode, India
 IEEE - 31661

IV. EXISTING MODEL scheduling virtual machines to Hosts[14] based on priority in

The existing algorithm for allocation of Cloudlets to VMs cloud. Our work focuses on scheduling tasks to the virtual
is FCFS(First Come First Served)[12]. The tasks are allocated machines based on priority. When the cloud user submits the
to the virtual machines in the order in which they arrive. tasks, the tasks are prioritized. Based on the priority, the tasks
FCFS Virtual Machine allocation policy works as follows: are allocated to the virtual machines. The task with highest
priority is assigned to the virtual machine with largest
 Cloud users submit Cloudlets and request for Virtual processing power. The steps are as follows:
machines.
 Cloud users submit Cloudlets, along with priority and
 The Cloudlet list and the VM request list are passed on request for Virtual machines.
to the Datacenter Broker.
 The Cloudlet list and the VM request list are passed on
 The DatacenterBroker uses CloudletScheduler to to the DatacenterBroker.
schedule Cloudlets.
 The Cloudlet list is sorted such that the Cloudlet of
 If this VM is free, Cloudlet will be sent to this VM. highest priority appears first.
 If not, Cloudlet will be sent to the next free VM.  The VM request list is sorted in decreasing order of the
processing power.
 The VM list is then passed on to the
VMAllocationPolicy for placement of VM on the Host.  The DatacenterBroker uses CloudletScheduler to
schedule Cloudlets.
 The Hosts will be created on Datacenters.
 The cloudlets are allocated to the VMs based on
Fig. 4 shows the flowchart for allocation of cloudlets to VMs
priority.
using First Come First Served algorithm.
 The Cloudlet with highest priority is assigned to the
VM with largest processing power.
 The VM list is then passed to VMAllocationPolicy for
the placement of VM on the Host.
 The Hosts will be created on Datacenters.
Fig. 5 shows the flowchart for allocation of cloudlets to VMs
using the proposed algorithm.

Fig. 4. Flowchart describing allocation of cloudlets to VMs using First Come

First served algorithm

The main drawback of the existing model is that the

Cloudlets are assigned to the VMs[13] as and when they
arrive. This increases the completion time of the jobs. The
user cannot prioritize the tasks for execution in the Virtual
machines. These drawbacks can be overcome by the proposed
model.
V. PROPOSED MODEL
Fig. 5. Flowchart describing allocation of cloudlets to VMs using proposed
In the proposed model, Cloudlets are assigned to VMs algorithm
based on priority. Some work have already been done for

4th ICCCNT 2013

July 4-6, 2013, Tiruchengode, India
 IEEE - 31661

A. Algorithm for allocating Cloudlets to VMs using the

proposed model

Input: Cloudlet list and VM request list

Output: Cloudlets mapped to VMs.

Get the cloudlets along with priority.

for each cloudlet in the cloudletlist do

begin
if this.cloudletpriority<next.cloudletpriority then
store this.cloudletpriority in a temporary variable
swap cloudletpriority
end
if this.propower<next.propower then
store this.propower in a temporary variable
swap propower
Fig. 6. Comparison of execution time of cloudlets using existing and proposed
end models
end
Fig. 6 shows the comparison of execution time of cloudlets
Map the cloudlets to corresponding VMs. using existing and proposed models. It is seen that if the
VI. SIMULATION RESULTS Cloudlets are assigned to VMs using the proposed model, the
execution time is very less.
The proposed model is simulated in CloudSim using java
programming language. The major benefit of simulation- VII. CONCLUSION AND FUTURE WORK
based approach is that it allows cloud developers to test the Cloud is a collection of interconnected and virtualized
performance of various scheduling policies in a controllable computer resources that are dynamically provided, based on
environment, before these policies are actually deployed in service level agreements, established by negotiation between
real cloud. This enables the developers to rectify their faults at service providers and customers. It allows workloads to be
an earlier stage. The existing and the proposed algorithms are deployed by the creation of virtual machines on
tested for their efficiency by creating fifteen Virtual Machines Hosts(physical machines) which are located at remote places.
which run on six Hosts. Each VM and Host is given specific Allocation of available resources[15] to user tasks is an
characteristics like processing capability, storage, bandwidth important issue. In this paper, the problem of scheduling of
and so on. The Hosts run on two Datacenters. The algorithms tasks[16] to virtual machines has been addressed. The cloud
are tested by varying the number of Cloudlets. In the proposed user prioritizes the tasks. The tasks are sent to the VMs based
algorithm, each cloudlet is given a specific priority. Based on on priority. The simulation results show that by using the
the priority, the cloudlet will be allocated to the virtual proposed system, there is a decrease in the execution time.
machines. The execution time of the Cloudlets is found using This decrease in execution time would in turn benefit the
the existing model and the proposed model. The results are cloud users. In future, we plan to implement this algorithm in
tabulated in Table I. a real-time cloud environment so that the cloud users would
benefit from decreased execution time. We also plan to
TABLE I. EXECUTION TIME OF CLOUDLETS
implement new scheduling policies for efficient allocation of
Number of Execution Time
Hosts to Datacenters.
Cloudlets (seconds)

Existing model Proposed model REFERENCES

[1] Michael Armbrust, Armando Fox, Rean Griffith, Anthony D. Joseph,
25 14.12 11.2 Randy H. Katz, Andrew Konwinski, Gunho Lee, David A. Patterson,
Ariel Rabkin, Ion Stoica and Matei Zaharia, “Above the Clouds: A
Berkeley View of Cloud Computing”, Technical Report No.
50 19.31 14.67 UCB/EECS-2009-28, February 10, 2009.
[2] Brian de Haaff, “Cloud computing -- the jargon is back!”, Cloud
Computing Journal, August 2008.
75 26.17 17.61
[3] R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, and I. Brandic.” Cloud
Computing and Emerging IT Platforms: Vision, Hype, and Reality for
100 32.77 20.68 Delivering Computing as the 5th Utility”, Future Generation Computer
Systems, 25(6):599-616, Elsevier, June 2009

4th ICCCNT 2013

July 4-6, 2013, Tiruchengode, India
 IEEE - 31661

[4] Siva Theja Maguluri and R. Srikant, Lei Yeng, “Stochastic Models of of Cloud Computing Infrastructures and Services”, Technical Report,
Load Balancing and Scheduling in Cloud Computing GRIDS-TR-2009-1, Grid Computing and Distributed Systems
Clusters”,published in proceedings of IEEE INFOCOM 2012. Laboratory, The University of Melbourne, Australia, March 13, 2009.
[5] Darshan Upadhyay, “Scheduler in cloud computing using open source [12] Rajkumar Buyya, Rajiv Ranjan and Rodrigo N. Calheiros, “Modeling
technologies”, Int. J. Computer Technology & applications, vol 3, 1093- and Simulation of Scalable Cloud Computing Environments and the
1098. CloudSim Toolkit: Challenges and Opportunities”, Proceedings of the
[6] Rodrigo N. Calheiros, Rajiv Ranjan, Anton Beloglazov, Cesar A. F. De 7th High Performance Computing and Simulation (HPCS 2009)
Rose, and Rajkumar Buyya, “CloudSim: A Toolkit for Modeling and Conference, Leipzig, Germany, June 21 - 24, 2009.
Simulation of Cloud Computing Environments and Evaluation of [13] Muhammad Shiraz, Abdullah Gani, “ Mobile Cloud Computing: Critical
Resource Provisioning Algorithms”, Software: Practice and Experience Analysis of Application Deployment in Virtual Machines”, International
(SPE), Volume 41, Number 1, Pages: 23-50, ISSN: 0038-0644, Wiley Conference on Information and Computer Networks (ICICN 2012)
Press, New York, USA, January, 2011. IPCSIT vol. 27 (2012) , Singapore.
[7] Archie Hendryx, “Cloudy Concepts: IaaS, PaaS, SaaS, MaaS, CaaS & [14] Subramanian S, Nitish Krishna G, Kiran Kumar M, Sreesh P, GR
XaaS”, Virutalization journal on November 1, 2011. Karpagam, “An adaptive algorithm for dynamic priority based virtual
[8] Gilad Parann-Nissany, “Cloud Compliance in IaaS Is Mainly Your machine scheduling in cloud”, International Journal of Computer
Responsibility”, Virtualization journal, February 15, 2013. Science Issues, Vol.9, Issue 6, No 2, November 2012.
[9] Gurudatt Kulkarni, Ramesh Sutar, Jayant Gambhir, “Cloud computing- [15] Amit Nathani, “Policy based resource allocation in IaaS cloud”, Future
Infrastructure as service- Amazon EC2”, International journal of Generation Computing Sysetms, Volume 28, issue 1, January 2012.
Engineering research and applications, volume2, issue 1,February 2012. [16] Sandeep Tayal, “Tasks scheduling optimization for the cloud computing
[10] Qi Zhang, Lu Cheng, Raouf Boutaba, “ Cloud computing: state-of-the- systems”,International Journal of advanced engineering sciences and
art and research challenges”, J Internet Serv Appl (2010) 1: 7–18, DOI technologies, vol.5, issue-2,111-115.
10.1007/s13174-010-0007-6, Published online: 20thApril 2010.
[11] Rodrigo N. Calheiros, Rajiv Ranjan, Cesar A. F. De Rose, and Rajkumar
Buyya, “CloudSim: A Novel Framework for Modeling and Simulation

4th ICCCNT 2013

July 4-6, 2013, Tiruchengode, India