Skip to Main Content
In this paper, we evaluate and compare the performance of several spanning tree routing strategies for divisible load scheduling on arbitrary graphs and derive recommendations as to which routing strategy provides a better trade-off between complexity and time performance. We consider a network comprising heterogeneous processors interconnected by heterogeneous links in an arbitrary manner. We evaluate the performance over a wide range of arbitrary dense graphs with varying connectivity and processor densities and study the effect of network scalability. In addition, we introduce a novel spanning tree routing strategy, which is referred to as minimum equivalent network spanning tree (EST), and analyze its performance. We apply the resource-aware optimal load distribution with optimal sequencing (RAOLD-OS) scheduling algorithm presented in the literature for obtaining an optimal solution. This study attempts to pool all known and applicable divisible load scheduling algorithms for arbitrary networks and presents a collective and comparative view of their performance.
Date of Conference: 16-19 Dec. 2009