Packet prediction for speculative cut-through switching

Abstract

The amount of intelligent packet processing in an Ethernet switch continues to grow, in order to support of embedded applications such as network security, load balancing and quality of service assurance. This increased packet processing is contributing to greater per-packet latency through the switch. In addition, there is a growing interest in using Ethernet switches in low latency environments such as high-performance clusters, storage area networks and real-time media distribution. In this paper we propose Packet Prediction for Speculative Cut-through Switching (PPSCS), a novel approach to reducing the latency of modern Ethernet switches without sacrificing feature rich policy-based forwarding enabled by deep packet inspection. PPSCS exploits the temporal nature of network communications to predict the flow classification of incoming packets and begin the speculative forwarding of packets before complex lookup operations are complete. Simulation studies using actual network traces indicate that correct prediction rates of up to 97% are achievable using only a small amount of prediction circuitry per port. These studies also indicate that PPSCS can reduce the latency in traditional store-and-forward switches by nearly a factor of 8, and reduce the latency of cut-through switches by a factor of 3.