A fast adaptive load balancing method for parallel particle-based simulations

DL Zhang and CJ Jiang and S Li, SIMULATION MODELLING PRACTICE AND THEORY, 17, 1032-1042 (2009).

DOI: 10.1016/j.simpat.2009.03.003

Balancing the work load can improve the performance of distributed simulation systems. In this paper we propose a fast adaptive balancing method, in which a binary tree structure is used to partition the simulation region into sub-domains. From a global view to local views, we balance the loads between sub-domains recursively by compressing and stretching sub-domains in group. This method can adjust the sub-domains with heavy loads and decompose their loads very fast. Then we compare the algorithm with two previously proposed algorithms by an artificial case and a real distributed case respectively. In both cases, our method can get a faster convergence speed and a lower communication overhead. (C) 2009 Elsevier B.V. All rights reserved.

Return to Publications page