TY - JOUR
T1 - An effective approximation algorithm for the Malleable Parallel Task Scheduling problem
AU - Fan, Liya
AU - Zhang, Fa
AU - Wang, Gongming
AU - Liu, Zhiyong
PY - 2012/5
Y1 - 2012/5
N2 - The Malleable Parallel Task Scheduling problem (MPTS) is an extension of one of the most classic scheduling problems (P∥C max). The only difference is that for MPTS, each task can be processed simultaneously by more than one processor. Such flexibility could dramatically reduce the makespan, but greatly increase the difficulty for solving the problem. By carefully analyzing some existing algorithms for MPTS, we find each of them suitable for some specific cases, but none is effective enough for all cases. Based on such observations, we introduce some optimization algorithms and improving techniques for MPTS, with their performance analyzed in theory. Combining these optimization algorithms and improving techniques gives rise to our novel scheduling algorithm OCM (Optimizations Combined for MPTS), a 2-approximation algorithm for MPTS. Extensive simulations on random datasets and SPLASH-2 benchmark reveal that for all cases, schedules produced by OCM have smaller makespans, compared with other existing algorithms.
AB - The Malleable Parallel Task Scheduling problem (MPTS) is an extension of one of the most classic scheduling problems (P∥C max). The only difference is that for MPTS, each task can be processed simultaneously by more than one processor. Such flexibility could dramatically reduce the makespan, but greatly increase the difficulty for solving the problem. By carefully analyzing some existing algorithms for MPTS, we find each of them suitable for some specific cases, but none is effective enough for all cases. Based on such observations, we introduce some optimization algorithms and improving techniques for MPTS, with their performance analyzed in theory. Combining these optimization algorithms and improving techniques gives rise to our novel scheduling algorithm OCM (Optimizations Combined for MPTS), a 2-approximation algorithm for MPTS. Extensive simulations on random datasets and SPLASH-2 benchmark reveal that for all cases, schedules produced by OCM have smaller makespans, compared with other existing algorithms.
KW - Approximation algorithm
KW - Parallel computing
KW - Scheduling algorithm
UR - http://www.scopus.com/inward/record.url?scp=84862784973&partnerID=8YFLogxK
U2 - 10.1016/j.jpdc.2012.01.011
DO - 10.1016/j.jpdc.2012.01.011
M3 - Article
AN - SCOPUS:84862784973
SN - 0743-7315
VL - 72
SP - 693
EP - 704
JO - Journal of Parallel and Distributed Computing
JF - Journal of Parallel and Distributed Computing
IS - 5
ER -