DocumentCode :
2224468
Title :
Performance evaluation of load distribution strategies in parallel branch and bound computations
Author :
Xu, Chengzhong ; Tschöke, Stefan ; Monien, Burkhard
Author_Institution :
Dept. of Electr. & Comput. Eng., Wayne State Univ., Detroit, MI, USA
fYear :
1995
fDate :
25-28 Oct 1995
Firstpage :
402
Lastpage :
405
Abstract :
Load distribution is essential for efficient use of available processors in a parallel branch-and-bound computation because the computation generates and consumes non-uniform subproblems at runtime. This paper presents six decentralized load distribution strategies. They are incorporated in a runtime support system, and evaluated in the solution of set partitioning problems on two parallel computer systems. It is observed that local averaging strategies outperform the randomized allocation and the Acwn algorithm significantly in large scale systems. They lead to an almost linear speedup in a PowerPC-based system with up to 32 nodes and to a speedup of 146.8 in a transputer-based system with 256 nodes. It is also observed that the randomized allocation and the Acwn algorithm can be improved by 10% to 15% when the bound information of subproblems is accounted for in the decision-making for load balancing
Keywords :
parallel algorithms; performance evaluation; resource allocation; Acwn algorithm; PowerPC-based system; decentralized load distribution strategies; decision-making; load balancing; load distribution strategies; local averaging strategies; nonuniform subproblems; parallel branch and bound computations; performance evaluation; runtime support system; set partitioning; transputer-based system; Computer science; Concurrent computing; Distributed computing; Distribution strategy; Large-scale systems; Libraries; Load management; Mathematics; Partitioning algorithms; Runtime;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Parallel and Distributed Processing, 1995. Proceedings. Seventh IEEE Symposium on
Conference_Location :
San Antonio, TX
ISSN :
1063-6374
Print_ISBN :
0-81867195-5
Type :
conf
DOI :
10.1109/SPDP.1995.530711
Filename :
530711
Link To Document :
بازگشت