On Batch-Processing Based Coded Computing for Heterogeneous Distributed Computing Systems
share
In recent years, coded distributed computing (CDC) has attracted significant attention, because it can efficiently facilitate many delay-sensitive computation tasks against unexpected latencies in different distributed computing systems. Despite such a salient feature, there are still many design challenges and opportunities. In this paper, we focus on practical computing systems with heterogeneous computing resources, and propose a novel CDC approach, called batch-processing based coded computing (BPCC), which exploits the fact that every computing node can obtain some coded results before it completes the whole task. To this end, we first describe the main idea of the BPCC framework, and then formulate an optimization problem for BPCC to minimize the task completion time by configuring the computation load and number of batches assigned to each computing node. Based on whether batch-induced overhead can be neglected or not, we develop two BPCC schemes, namely BPCC-1 and BPCC-2, for negligible and linear batching overheads, respectively. Through solid theoretical analyses, extensive simulation studies, and comprehensive real experiments on two heterogeneous distributed computing systems: 1) an Amazon EC2 computing cluster, and 2) an unmanned aerial vehicle (UAV)-based airborne computing platform, we demonstrate the high computational and energy efficiency of the proposed BPCC schemes.
READ FULL TEXT