SFS: Server-based flow scheduling for minimizing datacenter flow completion times

Abstract

Minimizing flow completion times (FCT) is a critical issue in datacenter networks. Existing approaches either fail to minimize FCT (i.e., DCTCP) or are costly to deploy (i.e., pFabric). This paper presents a server-based flow scheduling (SFS) scheme for enabling easy and rapid deployment in servers while almost retaining the same minimal FCT as state-of-the-art pFabric. To achieve the goal, SFS uses three novel techniques: receiver-driven flow scheduling, sender-side flow scheduling, and most recently seen flow coordination. With SFS, each sender or receiver keeps the highest-priority flow active and pauses the other low-priority ones, such that flows are completed one by one for minimizing FCT. Experimental results on NS2 show that SFS outperforms DCTCP and approaches pFabric in FCT.