On Wed, 05 May 2010 16:42:49 +0200, Georg Höllrigl <georg.hoellrigl@xxxxxxxxxx> wrote: > Hello, > > Can anyone explain me in some more detail what the elapsed time in the > squid access log exactly means? > > I found some definitions: > > duration The elapsed time considers how many milliseconds the transaction > busied the cache. It > differs in interpretation between TCP and UDP: > For HTTP this is basically the time from having received the request to > when Squid finishes sending > the last byte of the response. > > The most important question is, what may keeps squid so busy, that I get > sometimes numbers higher > thant 1000 ms? With reverse proxies I see such answers even with cache > hits and relative small files.... I think usually DNS lag, followed by TTL to the remote server for reads/writes, checking ACL lists (helpers and regex!), Disk I/O swapping, etc. Other big requests happening in parallel and flooding the event queues with I/O can also have some speed impact. The Measurement Factory and sponsors have added some extra DNS and timeout logging metrics to the most recent Squid-3 to display how much time is spent in the remote-delay areas. So admin can see how much is local delay and how much is unavoidably added by remote systems. The new metrics about which parts of HTTP request sequence get timed may be available in Squid-2.7, I'm not sure if they got accepted in from the 2.HEAD staging code. Amos
