A decentralized task scheduling algorithm and its performance modeling for computer networks

A dynamic taskschedulingalgorithm,that isstable,decentralized,and adaptivetonetwork topology,ispresented. Theproposedalgorithmisanextensionofnearestneighbor loadbalancingstrategywithanenhanceddegreeofefJiciencyand itisintendedformulticomputersconnectedbyastore andforward communicationnetwork. Theproposedalgorithm ismodeledby a central server openqueuingnetwork. It is shown that the responsetime of a task consistsof two parts. Thefirstpart comprisesa task‘ssettling time which consistsofscheduling time,communication time,and waiting timeinscheduling and communicationqueues. Thesecond part compriseswaiting time in the executionqueue in theexecutiontimeitself.Inordertoreducethefirstresponse time,thescheduling algorithmneeds tobe stable,so that a taskisquicklysettledat somenode. Onthe otherhand,the second response time isreduced if the algorithmejiciently migratesthe task toa lightlyloadednode.Theproposed algorithmis comprehensivelyevaluated,throughsimulation and analyticalmodel,and issh...

• Type: Conference
• Year: 1991
• Where: SPDP
• Authors: Ishfaq Ahmad, Arif Ghafoor, Kishan G. Mehrotra
• Keywords: Comprisesa Task‘ssettling Time, Distributed And Parallel Computing, Migratesthe Task Toa, SPDP 1991, Task Wander Cost