Beruflich Dokumente
Kultur Dokumente
Cloud computing is a controlled model that defines computing services, in which data
and resources are retrieved from the cloud service provider through the internet with the help
of web-based tools and functions. It is a collection of resources and services shared together Commented [R1]: Medium Plagiarism detected. This document
needs selective improvement.
and are provided to the users on pay-as-you-go model. An efficient task scheduling algorithm
is required, for mapping the resources with tasks. A Heuristic based algorithm is used to
attain the optimal or near optimal solution of task scheduling in the cloud environment. In
this paper, various types of task scheduling algorithms in cloud computing have been
discussed. Commented [R2]: Same as comment R1
Keywords
Cloud Computing, Task Scheduling, Heuristic, Optimal Solution.
1. INTRODUCTION
Cloud computing is a paradigm that facilitates convenient, ubiquitous, on-demand
broad n/w access to configurable computing resources that can be immediately provisioned Commented [R3]: Replace n/w by network
and released with minimum management effort [{http – A Review paper – 1}. It is the
distributed computing in which the tasks to resources will be distributed in some way to
facilitate communication and sharing the available resources.
tasks in the cloud environment. The aim is to reduce the makespan and increase the Commented [R5]: Tasks are executed with available resources.
How resources execute the scheduling task.
utilization of resources.There are three services associated with Cloud computing. They are
Infrastructure as a Service (IaaS), Platform as a Service (PaaS) and Software as a Service. Commented [R6]: SaaS could be included
These services are available in pay per use on demand. Cloud computing provides on-demand
accessing of the shared resources via the internet.
Scheduling is one of the important features of cloud computing. A fair scheduling will
improve the performance of the cloud system. Scheduling is mapping the tasks to the Commented [R7]: Give Justification. This sentence could be
written as “A fair scheduling improves ....
available and selected resources based on requirements. The overall performance should be
enhanced by reducing the completion time of all the tasks. Commented [R8]: Overall performance of what
2. TYPES OF SCHEDULING
specific and Grid computing environment information should be embedded in the scheduler. Commented [R12]: Not understandable
The local scheduling policy determines how the processes resident on a single CPU
are allocated and executed. The global scheduling policy uses information about the system
to allocate processes to multiple processors to optimize a system-wide performance. Cloud
scheduling falls into the global scheduling policy.
In static scheduling, information about all resources in the Cloud as well as all the
tasks in an application is assumed to be available by the time the application is scheduled. In
the case of dynamic scheduling, the basic idea is to perform task allocation on the fly as the
application executes [El-R 94]. Both static and dynamic scheduling is widely adopted in
Cloud Scheduling.
are
i. Makespan
ii. Resource Utilization
Makespan
Makespan is defined as the maximum completion time of resources [Geo 10]. Makespan is
calculated as,
Makespan = max(CT(ti, mj))
CTij = Eij + Rj
whereRj - Ready time of resource j after completing the assigned tasks
CT – Completion time of resources
Eij – Expected Execution time of task ‘i’ on resource ‘j’. Commented [R14]: What is tj and mj ? Check the formula.
Resource Utilization
Resource Utilization is defined as the amount of resources busy in executing tasks.
Resource Utilization = Mi * 100 / TARU. Commented [R15]: Explain
Max-Min
The Max-Min algorithm is just the reverse of the Min-Min algorithm. The Max-Min
heuristic begins with a set of all unmapped tasks. Then, find the set of minimum completion
times for each task. The task with overall maximum completion time from the set is selected
and assigned to the corresponding resource (hence said to be Max-Min). The mapped task is
removed from the set and the process repeats until all tasks are mapped.
MET
Minimum Execution Time (MET) assigns each task, in arbitrary order to the machine
with the best expected execution time for that task, regardless of that machine’s availability.
This algorithm is very simple and inexpensive but the loads among resources are not
balanced.
MCT
Minimum Completion Time (MCT) assigns each task in an arbitrary order to the
machine with the expected minimum completion time. It follows arbitrary order to map tasks
[Arm 98]. This causes some tasks to be assigned to resources that do not have the Minimum
Execution Time.
Sufferage
The Sufferage heuristic is based on the idea that better mappings can be generated by
assigning a machine to a task. The expected completion time would suffer more, if that
particular machine is not assigned to it. The sufferageheuristic sets priority to tasks based on
their sufferage value. It is defined as the difference between its best minimum completion
time (MCT) and its second MCT. Task with high sufferage value takes the precedence. By
assigning their first choice machines to the tasks that have the highest sufferagevalues among
all contending tasks, the Sufferage heuristic reduces the overall completion time.
4. EXPECTED TIME TO COMPUTE MODEL [Ali 00] Commented [R18]: Explanation of this model for including in
the related works.
In this model, it is assumed that the estimation or prediction of the computational load
of each task, the computing capacity of each resource, and an estimation of the prior load of
the resources are known in advance. Moreover, the Expected Time to Compute matrix (ETC)
of size m x n that is, ‘m’ number of tasks by ‘n’ number of resources, where each position
ETC[t][m] indicates the expected time to compute task t in resource m, is assumed to be
known or computable in this model. In the simplest of cases, the entries ETC[t][m] could be
computed by dividing the workload of task t by the computing capacity of resource m. This
formulation is usually feasible, since it is possible to know the computing capacity of
resources while the computation need of the tasks can be known from specifications provided
by the user, from historic data or from predictions [Hot 96]. Commented [R19]: Include references for the related work.
Check all are independent task scheduling algorithms.
Resource
Algorithm Makespan Advantages Issues
Utiliztion in % Commented [R21]: Describe all the algorithms in the related
works. OLB is not described. How you got the values for each
Implementation is
OLB 17 84.31 Poor makespan algorithm based the ETC matrix.
simple
Tasks may not be
Minimize the assigned to the best
MCT 15 64.58
makespan expected execution
time resource
Tasks are assigned
to the resources
MET 16 91.66 having best Severe load imbalance
expected execution
time
Tasks with longer
Minimize the
MIN-MIN 15 73.30 execution time is
makespan
penalized
Better mapping
compared to min- Tasks with shorter
min. execution time may
MAX-MIN 13 87.20
Better makespan have to wait for long
and reasonable time to execute
load balance
Communication cost,
Combination of
computation cost and
RASA 16 72.9 Min-min and Max-
deadline of each task
min
are not considered
6. CONCLUSION &FUTUTRE WORK
This paper presents the basic concepts of task scheduling in the Cloud
computing environment. Investigations are made on various factors that influence scheduling Commented [R22]: Name the factors
in a cloud system. This chapter also presents the scheduling policies and illustrates how they Commented [R23]: Replace chapter with paper
are combined with other ideas to provide an extensive work in the related area. It gives a deep Commented [R24]: It is not clear. Explain
understanding in order to identify the problem for the further research study. Commented [R25]: How?
From the above survey, new algorithm for independent task scheduling in cloud computing
can be developed. Commented [R26]: Investigated factors could be improved in
independent task scheduling for cloud environment.
REFERENCES