Load balancing in parallel battlefield management simulation on local- and shared-memory architectures
Abbreviated Journal Title
Comput. Syst. Sci. Eng.
load balancing; static and dynamic allocation; battlefield management; simulation; parallel time-driven simulation; Computer Science, Hardware & Architecture; Computer Science, Theory &; Methods
Load balancing is a critical issue for exploiting parallelism in any application and, particularly, in battlefield management simulation where the computational load dynamically changes with both time and space. Domain decomposition is an effective means to balance the load distribution in battlefield simulation. However, finer domain decompositions that lead to a better load balance incur heavier communication overhead. Earlier attempts at parallelizing battlefield simulation have traded load balance in favor of low communication overhead. We present three parallel battlefield management simulators, implemented on Intel's iPSC/2 and BBN's GP1000 multicomputers, with finer domain decomposition and address the communication overhead problem by processor allocation strategies that suit the underlying architecture of each machine. On the shared-memory GP1000, the strategy leads to a simulator with dynamic load balancing. Execution times of these simulators are provided, which show that the communication overhead is tolerable. Relevant details of the programs, data structures, and synchronisation mechanisms are included. The timing data collected have also been analyzed using Mathematica to obtain relative significance of the various simulator parameters and their interactions.
Computer Systems Science and Engineering
"Load balancing in parallel battlefield management simulation on local- and shared-memory architectures" (1998). Faculty Bibliography 1990s. 2218.