Brief Announcement

Abstract

Many data-parallel computing frameworks in today's datacenters consist of multiple computation and communication stages. A stage often cannot start or be completed unless all the required data pieces from the preceding stages are received. Coflow is a recently proposed networking abstraction to capture such communication patterns. We consider the problem of efficiently scheduling coflows with release dates in a shared datacenter network so as to minimize the total weighted completion time of coflows. This problem has been shown to be NP-complete, and several polynomial-time approximation algorithms have been recently proposed with provable performance guarantees. Our main result in this paper is a new polynomial-time approximation algorithm that improves the best prior known results. Specifically, we propose a deterministic algorithm with an approximation ratio of 5, which improves the prior best known ratio of 12. For the special case when all the coflows are released at time zero, we obtain an algorithm with an approximation ratio of $4$ which improves the prior best known ratio of 8.