FAMG: A flow-aware and mixed granularity method for load-balancing in data center networks

In current Data center Networks (DCNs), Equal-Cost Multipath (ECMP) is a default load-balancing scheme. However, using ECMP may result in the rapid growth of Flow Completion Time (FCT) due to its well-known drawbacks. In order to solve the problems of ECMP, load balancing schemes with per-packet granularity (such as RPS) and per-flowlet granularity (such as LetFlow) are proposed. These two different granularities have their features, but there is no existing load-balancing scheme considering both their advantages. Moreover, these schemes ignore the traffic characteristics in DCNs. In this paper, we mix per-flowlet and per-packet granularity in scheduling decisions and consider traffic characteristics. Based on this idea, we propose a flow-aware and mixed granularity method for load-balancing, referred to as FAMG. The key idea of FAMG is to distinguish elephant and mouse flows and schedule them with different granularities. We evaluate FAMG in simulations with NS-3. Experimental results show that FAMG slightly sacrifices elephant flows FCT but it is worth it since our results demonstrate that FAMG is a good trade-off that outperforms RPS, LetFlow, and TLB in terms of mouse flows FCT in both symmetric and asymmetric network topology.