DiFS: Distributed Flow Scheduling for Data Center Networks

Data center networks leverage multiple parallel paths connecting end host pairs to offer high bisection bandwidth for cluster computing applications. However, state of the art distributed multi-pathing protocols such as Equal Cost Multipath (ECMP) use static flow-to-link assignment, which is load-oblivious. They may cause bandwidth loss due to \emph{flow collisions} on a same link. Recently proposed centralized scheduling algorithm or host-based multi-pathing may suffer from scalability problems. In this paper, we present Distributed Flow Scheduling (DiFS) for data center networks, which is a switch-only distributed solution. DiFS allows switches cooperate to avoid over-utilized links and find available paths without centralized control. DiFS is scalable and can react quickly to dynamic traffic, because it is independently executed on switches and requires no synchronization. Extensive experiments show that the aggregate bisection bandwidth of DiFS using various traffic patterns is much better than that of ECMP, and is similar to or higher than that of a recent proposed centralized scheduling algorithm.