DocumentCode
3041051
Title
Average-case performance analysis and validation of online scheduling of independent parallel tasks
Author
Li, Keqin
Author_Institution
Dept. of Comput. Sci., State Univ. of New York, New Paltz, NY, USA
fYear
2004
fDate
26-30 April 2004
Firstpage
2
Abstract
Summary form only given. We analyze the average-case performance of an online scheduling algorithm for independent parallel tasks. We develop a method to calculate an analytical asymptotic average-case performance bound for arbitrary probability distribution of task sizes. In particular, we show that when task sizes are uniformly distributed in the range [1..C], an asymptotic average-case performance bound of M-(3-(1+1/C)C+1)C-1 can be achieved, where M is the number of processors. We also present extensive numerical and simulation data to demonstrate the accuracy of our analytical bound.
Keywords
performance evaluation; processor scheduling; program verification; asymptotic average-case performance analysis; independent parallel tasks; online scheduling algorithm; probability distribution; program verification; Algorithm design and analysis; Analytical models; Approximation algorithms; Computer science; Optimal scheduling; Performance analysis; Probability distribution; Processor scheduling; Random variables; Scheduling algorithm;
fLanguage
English
Publisher
ieee
Conference_Titel
Parallel and Distributed Processing Symposium, 2004. Proceedings. 18th International
Print_ISBN
0-7695-2132-0
Type
conf
DOI
10.1109/IPDPS.2004.1302899
Filename
1302899
Link To Document