Architectural Breakdown of End-to-End Latency in a TCP/IP Network

摘要

Adoption of the 10 GbE Ethernet standard as a high performance interconnect has been impeded by two important performance-oriented considerations: (1) processing requirements of common protocol stacks and (2) end-to-end latency. The overheads of typical software based protocol stacks on CPU utilization and throughput have been well evaluated in several recent studies. We focus on end-to-end latency and present a detailed characterization across typical server system hardware and software stack components. We demonstrate that application level end-to-end oneway latency with a 10GbE connection can be as low as 10μs for a single isolated request in a standard Linux network stack. The paper analyzes the components of the latency and discusses possible significant variations to the components under realistic conditions. We found that methods that optimize for throughput can significantly compromise Ethernet based latencies. Methods to pursue reducing the minimum latency and controlling the variations are presented.