Affinity Driven Distributed Scheduling Algorithm for Parallel Computations

摘要

With the advent of many-core architectures efficient scheduling of parallel computations for higher productivity and performance has become very important. Distributed scheduling of parallel computations on multiple places1 needs to follow affinity and deliver efficient space, time and message complexity. Simultaneous consideration of these factors makes affinity driven distributed scheduling particularly challenging. In this paper, we address this challenge by using a low time and message complexity mechanism for ensuring affinity and a randomized work-stealing mechanism within places for load balancing.rnThis paper presents an online algorithm for affinity driven distributed scheduling of multi-place2 parallel computations. Theoretical analysis of the expected and probabilistic lower and upper bounds on time and message complexity of this algorithm has been provided. On well known benchmarks, our algorithm demonstrates 16% to 30% performance gain as compared to Cilk [6] on multi-core Intel Xeon 5570 architecture. Further, detailed experimental analysis shows the scalability of our algorithm along with efficient space utilization. To the best of our knowledge, this is the first time affinity driven distributed scheduling algorithm has been designed and theoretically analyzed in a multi-place setup for many core architectures.