Performance Issues in Squid

Abstract

This paper describes the architecture of Squid proxy cache. It discusses the improvements in Squid achieved by three principles: reducing the file management overheads by using a large file, increasing CPU utilization for Squid with asynchronous I/O, and using scalable event notification facility, Kqueue, to improve proxy server under heavy loads. We implemented our proxy server based on the Squid according to the above three principles. From test results, we improve 390 req/sec to the version of Squid 2.3 stable 4. We also have about 310 req/sec improvement compared with the version of Squid 2.4 stable 6, the newest stable version now.