Issue No. 02 - February (1993 vol. 4)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/71.207592
<p>MULTIFIT-COM, a static task allocator which could be incorporated into an automatedcompiler/linker/loader for distributed processing systems, is presented. The allocator usesperformance information for the processes making up the system in order to determine anappropriate mapping of tasks onto processors. It uses several heuristic extensions of theMULTIFIT bin-packing algorithm to find an allocation that will offer a high systemthroughput, taking into account the expected execution and interprocessorcommunication requirements of the software on the given hardware architecture.Throughput is evaluated by an asymptotic bound for saturated conditions and under anassumption that only processing resources are required. A set of options are proposed foreach of the allocator's major steps. An evaluation was made on 680 small randomlygenerated examples. Using all the search options, an average performance difference ofjust over 1% was obtained. Using a carefully chosen small subset of only four options, afurther degradation of just over 1.5% was obtained. The allocator is also applied to adigital signal processing system consisting of 119 tasks to illustrate its clustering and loadbalancing properties on a large system.</p>
Index Termsdistributed systems; parallel systems; MULTIFIT-COM; static task allocator; automatedcompiler/linker/loader; distributed processing systems; bin-packing algorithm; high system throughput; interprocessor communication; asymptotic bound; digital signal processing system; clustering; load balancing; distributed processing; operations research; parallel processing; program compilers; resource allocation
G. Monforton and C. Woodside, "Fast Allocation of Processes in Distributed and Parallel Systems," in IEEE Transactions on Parallel & Distributed Systems, vol. 4, no. , pp. 164-174, 1993.