Congestion Control for Switched Ethernet

Interconnects for clusters and bladed systems must deliver efficient throughput, low latency, low delay variations and minimal frame drops. The primary technical issues hindering Ethernet adoption for cluster and blade system interconnects are the current methods Ethernet switches use for dealing with congestion, which can happen frequently under cluster and blade system workloads. The common response to congestion is to drop frames and the common method of avoiding the need to drop frames is to utilize very large switch buffers. In this paper, we propose the insertion of a simple self-managing congestion control protocol into existing communication stacks at the edges of layer 2 switched interconnects. We assert that control of the traffic flow into the layer 2 subnet is key to controlling the characteristics of cluster and blade system interconnects. We show simulation results demonstrating how the proposed protocol, coupled with a layer 2 ingress rate control function, can dynamically control traffic flow so as to maximize the throughput efficiency while minimizing the loss, delay, and delay variations.