Issue No. 09 - September (1993 vol. 4)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/71.243527
<p>Load balancing requirements in parallel image analysis are considered and results on theperformance of parallel implementations of two image feature extraction tasks on theConnection Machine and the iPSC/2 hypercube are reported and discussed. A loadredistribution algorithm, which makes use of parallel prefix operations and one-to-onepermutations among the processors, is described and has been used. The expectedimprovement in performance resulting from load balancing has been determinedanalytically and is compared to actual performance results obtained from the aboveimplementations. The analytical results demonstrate the specific dependence of theexpected improvement in performance on the computational and communicationrequirements of each task, characteristic machine parameters, a characterization of priorload distribution in terms of parameters which can be computed dynamically at the startof task execution, and the overhead incurred by load redistribution.</p>
Index Termsload balancing; parallel implementations; image feature extraction; performance;Connection Machine; iPSC/2 hypercube; load redistribution algorithm; parallel prefixoperations; one-to-one permutations; communication requirement; load redistribution;distributed memory systems; feature extraction; hypercube networks; image processing;parallel machines; performance evaluation; resource allocation
D. Gerogiannis and S. Orphanoudakis, "Load Balancing Requirements in Parallel Implementations of Image Feature Extraction Tasks," in IEEE Transactions on Parallel & Distributed Systems, vol. 4, no. , pp. 994-1013, 1993.