IJSRD - International Journal for Scientific Research & Development| Vol.

4, Issue 03, 2016 | ISSN (online): 2321-0613

Recent Advancements in Scheduling Algorithms in Cloud Computing
Ravi Kumar1 Anil Kr. Panghal2
Department of Computer Science and Engineering
H.C.T.M. Technical Campus (Kaithal), Kurukshetra University, Kurukshetra
Abstract— Now a days, cloud computing is most important
paradigm used by almost every online user directly or III. CLOUD SCHEDULING ALGORITHMS
indirectly. To meet the large number of users a lot of The following scheduling algorithms have been established
improvement has been required. For that reason, research in the area of grids and clouds.
work has been carried out time to time. In cloud, for each MSDoT- Memory size and depth of tier based
task user may use hundreds of thousands of resources, so queuing model for scheduling a multi-tier-cloud was
manual scheduling is not a feasible solution. This paper proposed by Stanislas , L. Arockiam [4]. The algorithm was
describes the brief and organized analysis on most recent designed to minimize resource utilization and response time.
cloud scheduling algorithms. To minimize resource utilization, virtualization technology
Key words: Cloud Computing, Makespan, Cost, Energy and was used in a cloud environment. The algorithm was based
Cloud Service Provider (CSP) on queuing models to schedule the requests in a vertically
scaled virtualized machine. The working of algorithm was
I. INTRODUCTION based on the properties of requests such as depth of tier and
Cloud Computing [1] is a technology that uses the internet memory size. Overall performance was improved by
and central remote servers to maintain data and applications. maintaining average response time. Non linear optimization
An example of widely used cloud computing is yahoo mail, model was used to minimize the number of virtual machines
Gmail or hotmail. The advantages of cloud computing are for the requirement of the customer.
flexibility, work from anywhere, document control, disaster A V. Karthick, Dr. E. Ramaraj and R. Kannan [5]
recovery, security and environment friendly. According to proposed Tri Queue Scheduling (TQS) algorithm TQS is
National Institute of standards and technology (NIST)[2] based on Dynamic Quantum Time. The most challenging
“Cloud computing is a model for enabling convenient, on- task in cloud computing is resource sharing. Jobs are
demand network access to a shared pool of configurable grouped into three different queues i.e. Small, medium and
computing resources (e.g., networks, servers storage, long based on time needed to allocate resources and the
applications, and services) that can be rapidly provisioned processor needed. Jobs are sorted in a descending order. In a
and released with minimal management effort or service long queue first 40% jobs are stored, in a medium queue
provider interaction”. next 40 % jobs and in a small queue remaining 20% jobs are
stored. The jobs are than handled by resource manager. TQS
II. CLOUD SCHEDULING reduces the fragmentation problem and starvation within the
Cloud scheduling [3] algorithms typically aim to minimize process. Round robin fashion was used to give equal
the total time and total cost for processing all tasks. importance to all jobs based on dynamic quantum time.
However, in cloud computing environments, computing Firstly, jobs submitted by the user are passed to the resource
capability differs for different resources and so does the cost manager. Resource Manager is responsible for queuing and
of resource usage. There are various types of cloud scheduling of jobs. Resource Manager is divided into job
scheduling. User level scheduling (comprises of market queue and job scheduler. Arrived jobs first updated in job
based and auction based scheduling like FIFO scheduling, queue and priority has been assigned by scheduler based on
priority based, non preemptive scheduling etc.).Cloud dynamic quantum time. In cloud environment,
service scheduling is classified at user level and system Metascheduler is responsible for managing resources. Three
level. At user level, it mainly considers the service regarding queues are formed long, medium and small. The jobs are
problems between the provider and the user. At system then sent to the cloud environment.
level, scheduling and resource management is done. Static Energy-aware scheduling scheme using workload-
scheduling permits pre-fetching of required data and aware consolidation technique in cloud data centres was
pipelining of different stages of task execution. Static proposed by Li Hongyou; Wang Jiang yong; Peng Jian;
scheduling imposes less runtime overhead. In case of Wang Junfeng and Liu Tang [6]. This paper consists of two
dynamic scheduling, information of the job components or algorithms to reduce energy consumption in cloud data
task is not known before. Thus the execution time of the centres. The two algorithms are the Energy-aware
task may not be known and the allocation of tasks is done Scheduling algorithm using Workload-aware Consolidation
only as the application executes. In cloud environment, Technique (ESWCT) and the Energy-aware Live Migration
heuristic based scheduling can be done for more optimized algorithm using Workload-aware Consolidation Technique
results. For the management of workflow execution, (ELMWCT). The two algorithms based on the fact that the
workflow scheduling is done. Real Time Scheduling in multiple resources (such as CPU, memory and network
cloud environment is done to increase the throughput and to bandwidth) are shared by users concurrently in cloud data
decrease the average response time instead of meeting centres and heterogeneous workloads have different
deadline. resource consumption characteristics. Both algorithms check
the problem of consolidating heterogeneous workloads.

All rights reserved by www.ijsrd.com 154
Recent Advancements in Scheduling Algorithms in Cloud Computing Environment
(IJSRD/Vol. 4/Issue 03/2016/042)

They try to execute all Virtual Machines (VMs) with the to the sorted list to the available machine and then PSO was
minimum amount of Physical Machines (PMs), and then applied.
power off unused physical servers to reduce power Xiaoli Wang , Yuping Wang and Yue Cui [10]
consumption. The results show that both algorithms proposed energy-aware bi-level optimization model for
efficiently utilise the resources in cloud data centres, and the multi-job scheduling problems under cloud computing. The
multidimensional resources have good balanced utilizations, model mainly focused on improving the energy efficiency of
which demonstrate their promising energy saving capability. servers through appropriate allocation of schemes and task-
M. Geethanjali et al. [7] proposed Multi-Objective Real scheduling strategies. The model describes the relationship
Time Scheduling Algorithm (MORSA). The algorithm was between the performance and energy consumption of
designed for Multi Cloud Environments. The objective servers. For the proper utilization of energy over the server,
behind the algorithm is to achieve minimum cost for the job map reduce technique has been used. Since, network
execution. CSPs provide false information in order to bandwidth is a scarce resource in cloud computing, the
achieve maximum profit. To get truthful information from model guarantees that each task runs on the same node that
CSPs, game theory is used. The algorithm also minimizes the split resides on that for achieving data locality. Task
the deadline constraints of the user’s task. Winner selection scheduling strategies depend directly on appropriate data
is done on the basis of minimum time and the cost placement policies; the model is formulated as an integer bi-
associated with the service provider that satisfies the level programming. For the efficient working of the model
deadline constraint. MORSA performs fast scheduling than encoding and decoding methods have been designed. GA
SPEA2 and NSGA-II. has been used for enhancement in searching ability.
Multi cloud environment contains several cloud Wen An Tan [11] proposed Trust oriented
coordinators (datacenters). The cloud coordinator scheduling algorithm. The algorithm adopts a trust metric
communicates, negotiates and exchange information from that combines both direct trust and recommendation trust. In
other coordinator of another cloud service provider addition, algorithm provides balance policies to enable users
regarding resources. Cloud coordinator provides quality of to balance different requirements, including time, cost, and
service according to the user’s budget. trust. Due to dynamic and failure prone large distributed
Chun-Yan Liu, Cheng-Ming Zou and Pei Wu [8] system lead to lack of trust between enterprise and provider
proposed task scheduling algorithm based on genetic which results that enterprise does not fully adopt cloud
algorithm and ant colony optimization to achieve optimal services. Cloud service selection algorithm had been
solution. The algorithm solves the slow convergence designed to create optimum workflow applications and to
problem caused by the lack of initial pheromone of ACO. meet user constraints. The algorithm is compared with
The algorithm is based on the integration of genetic MCP, greedy cost and DBL (Deadline Bottom level). MCP
algorithm and ACO (Ant Colony Optimization). Strong stands for minimum critical path. MCP sorts services
search capability of GA has been used to provide better according to the time and selects the service that has
solution. The solution is then converted into initial minimum time for the corresponding task. Despite of that,
pheromone of ACO. Positive feedback of ACO gives the greedy-cost approach sorts services according to the cost
optimal solution. The integration of two improves searching and selects the service with the minimum cost. MCP has the
efficiency of algorithm. This method consists of two stages shortest execution time among the algorithms but it has
i.e. Map and reduction. In the Map stage, MapReduce highest cost. Conversely, greedy cost has the least cost but it
function divides the user tasks into smaller sub tasks and has longest execution time. Lastly in the TWFS (Trust
allocated the sub tasks to multiple workers who generated Service Oriented Workflow Scheduling) approach, time and
the intermediate file. In reduce method; the result will be cost are both considered simultaneously, that enables the
generated after the research pooled analysis. Adaptive user to compromise requirements to yield a genuinely better
crossover mechanism and single point mutation is used in solution.
this algorithm. Eco aware algorithm was proposed by Xiang Deng,
Amandeep Verma and Sakshi Kaushal [9] Di Wu, Junfeng Shen, and Jian He [12]. Eco power
proposed Bi-Criteria Priority Based Particle Swarm algorithm has been designed for eco-aware power
Optimization Algorithm for Workflow Scheduling in Cloud. management and load scheduling for geographically
BPSO was designed to schedule workflow tasks over the distributed green cloud centers. The goal of the algorithm is
accessible cloud resources that minimized the execution to minimize the time-average eco-aware power cost and to
time and the execution cost under the user specified deadline meet the better quality of experience (QoE). Lyapunov
and budget constraints. This was simulated with four optimization framework has been used to get provable
different real world applications of workflow and performance bounds. Two load scheduling strategies has
comparison was done with BHEFT i.e. Budget Constrained been used i.e. Cheap-First Load Scheduling Strategy and
Heterogeneous Earliest Finish Time and standard PSO. The QoE-First Load Scheduling Strategy. Cheap- first load
Algorithm significantly decreased the execution cost of scheduling is used to minimize the total electricity cost
schedule as compared to BHEFT and PSO under the same incurred by drawing electricity from the power grid. QoE-
Deadline and Budget Constraint and using same pricing First load scheduling is used to minimize the total latency
model. Each workflow’s task was assigned priority using when users are requesting services. Online Eco power
bottom level. These priorities were then used to initialize the algorithm consists of two power management plans. In the
PSO. First step is modified in which bottom level was first plan different power sources can automatically make
calculated and according to bottom level all the tasks were switches based on preset thresholds. The second plan is the
sorted. After that tasks were assigned sequentially according

All rights reserved by www.ijsrd.com 155
Recent Advancements in Scheduling Algorithms in Cloud Computing Environment
(IJSRD/Vol. 4/Issue 03/2016/042)

variant of the online power control method. The method REFERENCES
does not consider the renewables. [1] http://www.wikinvest.com/concept/Cloud_Computing
Multi Queue Job Scheduling Algorithm was [2] Mell P., Grance T., “The NIST Definition of Cloud
proposed by A V Karthick, Dr. E. Ramraj and R. Ganapthy Computing”, NIST Special Publication 800-145,2011.
Subramanian [13]. The algorithm was designed to overcome [3] Chawla Y., and Bhonsle M., “A Study on Scheduling
the fragmentation problem and to reduce the starvation with Methods in Cloud Computing”, International Journal of
in the process. MQS is based on burst time and gives equal Emerging Trends & Technology in Computer Science
importance to all jobs. Jobs are selected dynamically and (IJETTCS), 2012, pp. 12-17.
utilize free unused space in an economic way. Jobs are [4] Stanislas A., and L. Arockiam L., “MSDoT- Memory
sorted in an ascending order. Three queues are formed Size and Depth of Tier Based Queuing Model for
i.e.small queue medium queue and long queue. Small queue Scheduling a Multi-Tier Cloud”, ICETACS, 2013, pp.
consists of short burst time and stores 40% jobs, Medium 119-124.
queue consists of medium burst time and stores another 40% [5] Karthick A.V., Ramaraj E. and Kannan R., “An
jobs and long queue consists of long burst time and stores Efficient Tri-Queue Job scheduling using Dynamic
next 20% jobs. The scheduler takes jobs from each queue. Quantum Time”, IEEE International Conference on
Jobs submitted by clients are sent to the Queue manager by Green Computing, Communication and Conversation of
service provider. Queue manager sorted all jobs in Energy,2013, pp. 871-876.
ascending order according to burst time and maintain three [6] Hongyou Li, J. Wang , J.Peng, J.Wang and T. Liu,
queues as shown in figure. The jobs are then sent to cloud “Energy-aware scheduling scheme using workload-
environment. aware consolidation technique in cloud data centres”,
Anti starvation algorithm was proposed by Heba IEEE, 2013, pp. 114-124.
Kurdi et al [14] for handling resource rejection and [7] Geethanjali M., Sujana M.J.A.J., Revathi T., “Ensuring
starvation problem in public clouds. The algorithm is an Truthfulness for Scheduling Multi-objective Real Time
enhancement to the Haizea scheduler which results in Tasks in Multi Cloud Environments”, IEEE
reduction of rejection rate and starvation problem and International Conference on Recent Trends in
increased the resource utilization. The algorithm, prevents Information Technology, 2014, pp. 1-7.
the AR (Advanced Reservations) leases from preempting [8] Liu C.Y., Zou C.M., Wu P., “A task scheduling
BE (Best Effort) leases on the basis of threshold if the leases algorithm based on genetic algorithm and ant colony
were preempted a certain number of times. The algorithm optimization in cloud computing”, IEEE 13th
uses a combination of aging and negotiation methods. Aging International Symposium on Distributed Computing
is used to compute the number of times each BE leases has and Applications to Business, Engineering and Science,
been suspended. As the aging counter increases, priority of 2014, pp. 68-72.
BE lease has been increased. Negotiation is used to prevent [9] Verma A., and Kaushal S., “Bi-Criteria Priority Based
the rejection of AR lease when resources are unavailable. Particle Swarm Optimization Workflow Scheduling
Negotiation method gives chance to customer to select from Algorithm for Cloud”, Proceedings of 2014 RAECS
rejecting AR lease and converting it to BE lease. UIET Panjab University Chandigarh, 2014, pp. 1-6.
Enhancing cloud computing reliability using [10] Wang X., Wang Y., Cui Y., “An energy-aware bi-level
efficient scheduling by providing reliability as a service was optimization model for multi-job scheduling problems
proposed by Abishi Chowdhury and Priyanka Tripathi [15]. under cloud computing”, 2014, Springer ISSN 1432-
This paper consists of a reliability computing technique to 7643.
calculate the reliability of a Cloud data centre. Further, this [11] Tan W., Yong Sun Y., Li L.X., Lu G., and Wang T.,“A
paper describes the mechanism for continuous updating of Trust Service-Oriented Scheduling Model for Workflow
cloud resources' reliability and providing a reliable Applications in Cloud Computing”, IEEE Systems
scheduling of the resources to the cloud users in a cloud Journal, VOL. 8, NO. 3,2014, pp. 868-878.
computing environment. [12] Deng A. X., Wu D., Shen J., and He J., “Eco-Aware
Multi Queue Genetic Algorithm .proposed by Neha Online Power Management and Load Scheduling for
et al [16]. The proposed Multi Queue Genetic Algorithm Green Cloud Datacenters” IEEE Systems Journal, 2014,
(MQGA) mainly focuses on minimizing the makespan. The pp. 1-10.
algorithm performs proper utilization of resources and [13] A.V. Karthick, Dr.E.Ramraj, R.Ganapthy Subramanian,
reduces starvation problem. The proposed Multi Queue “An Efficient Multi Queue Job Scheduling for Cloud
Genetic Algorithm (MQGA) requires the schedule generated Computing”, IEEE, World Congress on Computing and
by multi queue scheduler to further optimize the makespan Communication Technologies, 2014, pp. 164-166.
and efficient utilization of resources. Proper allocation of [14] Kurdia H., Aloboudb E., Alhassanb S., Alotaibib E.T.,”
resources results in the reduction of makespan. An Algorithm for Handling Starvation and Resource
Rejection in Public Clouds”, Procedia Computer
IV. CONCLUSION Science 34 ( 2014 ) , 2014, pp. 242 – 248.
This paper covers the most recent scheduling algorithms and [15] C. Abishi and T. Priyanka , “Enhancing cloud
in future work can be done on MQGA algorithm. The new computing reliability using efficient scheduling by
parameters like cost, time complexity can be introduced. providing reliability as a service”, IEEE, 2014, pp. 99-
[16] L.Neha et al. ,“MQGA: Evolutionary Method for Cloud
Task scheduling”,vol 5, issue 6, IJARCSSE, 2015.

All rights reserved by www.ijsrd.com 156

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master Your Semester with a Special Offer from Scribd & The New York Times

Cancel anytime.