Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

So looking over the logs the average response time is around 20-30ms but the load on the machine is 20+ and using an external test things seem to take over a second to load. I can't imagine this is all network latency so I'm curious at what point the apache logs start to track the request. If the load is at 20 does that mean things are waiting in line to even get to the acpache process but once they get there it only takes 30ms to process?

share|improve this question
up vote 6 down vote accepted

I assume you mean the response-time as delivered by %D in the log-line (i.e. time in microseconds to serve the request). The time stated there is the duration from the successful reading of the URI (for instance GET / HTTP/1.1) but before the rest of the headers to the logging of said request, which happens once it has been served completely (in source-code terms this would be from httpd:// to So yes, once Apache actually gets to read said request, it only takes 30ms.

Whether the added latency is due to the load of the machine depends on where that load is generated, and how your Apache httpd is configured. One explanation would be that all your httpd workers are busy and the request gets stuck in the accept-queue (see for a bit more technical background of what that is; short version is that httpd does not immediately have to accept a connection if there are no available workers).

To get an indication of whether this is happening would be to look at httpd's status-page (localhost/server-status, you will want to have mod_status ( configured for that) and look at whether you have any idle workers left (and whether they are "enough"). If not, try increasing the number of workers httpd spawns (the exact configuration of which depends on which MPM-module you use; for mpm_worker (, you would increase MaxClients, ThreadsPerChild, ServerLimit, etc. as appropriate. Take care not to exceed your machine's available memory or maximum of open files though (remember that each connection is at minimum one filehandle).

Depending on /why/ all the apache workers are tied up, this may help you. If they are all tied up in long-running dynamic page generation (i.e. mod_php or the like), all this is gonna do is increase the load on your system. If the load on your system is from a sub-par storage subsystem and all the Apache processes are tied up in large file transfers and waiting on the I/O subsystem, you'll just get more load. There are workarounds for both; if it's dynamic content, you will want to limit concurrent access to long-running scripts (for instance by getting PHP out of the worker process and into, say, a FastCGI or even CGI process and limiting the amount of those to somewhere below the MaxClients setting so that non-dynamic objects can be served simultaneously without incurring an accept-backlog on them.

Sorry for the missing http-links, apparently my reputation on this site is not enough to actually post links.

share|improve this answer

I'm not certain if this Apache installation is running on a Windows server but if so you could use a combination of perfmon (with Network Interface / Output Queue Length to report network queues, if any) and Firefox + Firebug to document response times. This could give you some insight into where the problem may lie.

I also found this article on mod_log_config, not sure if you're using this or another module for logging. I don't have a development Apache install to test it on myself but it might be worth a read.

Good luck.

share|improve this answer
The "Load of +20" makes me think this is a Linux system – Brent Nov 24 '09 at 14:47

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.