Anytime coding for distributed computation

Abstract

A novel coding scheme is proposed to speed up distributed computation through a form of approximate computing. It is known that task replication can greatly mitigate the “straggler effect” in cloud computing, wherein an overall computation can be significantly delayed by slowed processing nodes (or “stragglers”). It has also been demonstrated that, in certain contexts, ideas of error-correction coding can more efficiently deal with stragglers than pure replication. The approach proposed herein builds on these earlier observations through an “anytime” approach to approximate computing. In this paradigm, over time one can produces approximate solutions of increasing accuracy. To accomplish this we first decompose a computational job in to tasks of various priorities. Next, we apply linear error correction coding to produce subtasks that are assigned to different processors. The decomposition used has a big effect on the type of anytime performance we attain. We study this scheme in a general framework in terms of the expected cost of the approximate solution. We further explore the approach in the context of vector-matrix multiplication. The proposed construction is numerically studied and, in comparison to previous work, demonstrates a significant improvement in the accuracy/latency trade-off.