Reconciling mice and elephants in data center networks

Small switch buffers, high-speed links, short round-trip times and the composite nature of the traffic in data center networks (DCN) lead to several congestion problems that are not handled well by traditional congestion control mechanisms such as TCP. In this paper we design a simple switch-driven, flow-aware, congestion control algorithm to deal with such congestion issues. The basic idea of the proposed mechanism is reminiscent of classic congestion control in flow-aware networks such as ATM-ABR, where the switch sets a field in packet headers to enforce the sending rate at the source; less classic though are the challenges faced in designing such flow-awareness in the flow-aversive IP environment without modifying the TCP sender and receiver algorithm, to enable deployment in public data centers. We discuss in this paper our algorithm and give numerical results from NS-2 simulations to show its effectiveness in achieving high throughput overall, a good fairness and short flow completion times for delay-sensitive flows¹.