On Thu, 2008-03-27 at 12:21 +1100, Adam Carter wrote: > Following on from my comment above, a single 20gig RAID0 cache_dir is > probably not that much different to two 10gig cache_dirs on single > disks. If using aufs then the RAID0 would only run as a single thread > so that may adversely affect performance. No, the minimum number of threads is 16, and increasing slightly with the number of cache_dir. > I guess that RAID0 would offer a worse seek time from squid's > perspectiv as each request from squid is serialised, but data transfer > rate will be higher for a particular object. Negliable on both. RAID0 transfer rates only becomes visible on very large batch requests, which is very rarely seen with Squid.. > I imagine squid is more sensitive to seek than throughput. I'm just > speculating on all this... Correct. Squid eats a lot of seek time, and fairly little bandwidth. > Do you know what your "600 - 700 Req/Sec Polygraph polymix-4 > benchmark" is in Mbps? Comparing Mbps with a simulation isn't that relevant. Req/Sec is a much more reliable measure when comparing with real life. Regards Henrik
