Adrian Chadd wrote: > Hi everyone, > > The COSS code in Squid-2.6 has come quite far from its original design by > Eric Stern. Steven Wilton has put an enormous amount of effort into the > COSS design to fix the remaining bugs and dramatically improve its > performance. > > I've assembled a quick webpage showing the drop in CPU usage and the > negligible effect on hit-rate. Steven Wilton provided the statistics > from two Squid caches he administers. > > You can find it here - http://www.squid-cache.org/~adrian/coss/. > Steven is running a recent snapshot of squid-2.6. The latest -STABLE > release of Squid-2.6 doesn't incorporate all of the COSS bugfixes > (and there's at least one really nasty bug!) so if you're interested > in trying COSS out please grab the latest Squid-2.6 snapshot from > the website.
The example proxy given has a request rate of about 100 req/s max, if I understand the graphs correctly. How does COSS hold when the request rate is significantly higher? I run a proxy that currently seems to peak around 420 req/s (and has an average rate of about 300 req/s during office hours), and am currently using aufs. Mbps peakrate is about 25/30 Mbps. Anything that can improve the proxy performance even more is wanted, since I have the feeling that currently the proxy is hitting its upper limits. Joost