Midjfr Global Scheduling Algorithm for Heterogeneous Grid

Only available on StudyMode
  • Topic: Grid computing, Scheduling, Computational complexity theory
  • Pages : 11 (2932 words )
  • Download(s) : 115
  • Published : January 26, 2013
Open Document
Text Preview
Letter Paper
© 2011 ACEEE
DOI: 01.IJRTET.06.01.
Int. J. on Recent Trends in Engineering and Technology, Vol. 6, No. 1, Nov 2011 606
MidJFR Global Scheduling Algorithm For
Heterogeneous Grid Environment
Dr.G.Sumathi1, R.Santhosh Kumar2, and S.Sathyanarayanan3
Sri Venkateswara College of Engineering, Sriperumbudur, India Email :{ sumathiganesan, santhosh.11292,sathyanarayanan}@gmail.com Abstract--Grid computing is now an available option to
traditional supercomputing environments. Computational
grid is one of the variants of grid which provides parallel
computing using available resources to do major
computational tasks organised modular wise over different or same geographical area. Scheduling in grid is the process of assigning parallel resources to carry out a task. Priority Based SchedulingAlgorithm is utilised in this paper. The proposed

MidJFR algorithm is a revised version of the priority based
algorithm. Working of this algorithm is based on the parameter ‘priority’. MidJFR refers to Medium Job Fastest Resource algorithm which generally classifies the jobs into three tier categories, High, Middle and Low based on their priority .In MidJFR algorithm the job of medium computational

complexity and resources exhibiting medium level of
parallelism are assigned with a high priority. The value for level of parallelism is assigned based on the amount of
parallelism exhibited by the job and the amount of parallelism exhibited by the available resources. The effectiveness of this algorithm is evaluated through simulation results.
Index Terms--Computational Grid, Global scheduling,
Computational Complexity, Level of Parallelism, Directed
Acyclic Graph (DAG)
Computational Grid is one that performs any form of
calculation. It provides high end resources for executing
parallel tasks. The resources available in the grid could be a single processor, a symmetric multiprocessor cluster, a
distributed memory multiprocessor system, or a massively
parallel supercomputer. A local scheduler is different from
grid scheduler. A grid scheduler is in charge of resource
discovery, resource allocation and job execution management. While a local scheduler manages a cluster also owning the
resource. A proper scheduling and efficient load balancing
can lead to improved overall system performance in
heterogeneous grid environment with its multiple sources,
also leading to a lower turn-around time for individual
jobs.[1]First In First Out (FIFO) algorithm neither considers any of the job parameters nor the resource parameters. This
proposed MidJFR algorithm is a revised version of the priority based algorithm. The SJFR and LJFR algorithms consider
computational complexity of jobs for scheduling and ignore
the priority of a job.
The priority based algorithm assigns the job with a high
priority based on the factors like computational complexity
and the level of parallelism exhibited. The proposed MidJFR
algorithm also works based on the same parameters as that
of the priority based algorithm but outcomes all global
scheduling algorithms with its economical usage of resources and its consistent performance for the jobs assigned.
The rest of the paper is organized as follows.
The grid framework is presented in Section II, the proposed
scheduling algorithm is discussed in Section III , the
performance study is carried out and results are discussed in Section IV and finally, some concluding remarks in Section V. II. GENERAL FRAMEWORK OF A GRID
Fig.1 shows the framework of the grid .The grid is mainly
formed by the following components: The Global Grid
Resource Brokers (GGRB), Local Grid Resource Brokers
(LGRB) and Grid Information Server (GIS). [1]The above listed components have their independent functionalities that help
in grid management and job scheduling and thus serve the
purpose of a grid.
tracking img