Dependability of Scheduling between latency-sensitive small data flows (a.k.a. mice) and throughputorientedlarge ones (a.k.a. elephants) has become ever challenging with the proliferation of cloudbased applications. In light of this mounting problem, this work proposes a novel flow controlscheme, HOLMES (HOListic Mice-Elephants Stochastic), which offers a holistic view of globalcongestion awareness as well as a stochastic scheduler of mixed mice-elephants data flows in DataCenter Networks (DCNs). Firstly, we theoretically prove the necessity for partitioning DCN pathsinto sub-networks using a stochastic model. Secondly, the HOLMES architecture is proposed, whichadaptively partitions the available DCN paths into low-latency and high-throughput sub-networksvia a global congestion-aware scheduling mechanism. Based on the stochastic power-of-two-choicespolicy, the HOLMES scheduling mechanism acquires only a subset of the global congestioninformation, while achieves close to optimal load balance on each end-to-end DCN path. We alsoformally prove the stability of HOLMES flow scheduling algorithm. Thirdly, extensive simulationvalidates the effectiveness and dependability of HOLMES with select DCN topologies. The proposalhas been in test in an industry production environment.
展开▼
