Comments on: Some Tuning Tips For Apache Mod_Cache (mod_disk_cache)

By: indibest.com

indibest.com — Sat, 29 Oct 2011 09:16:35 +0000

Excellent work. I was able to reduce MaxClients after the changes

By: Dennis van der Heijden

Dennis van der Heijden — Sat, 30 Oct 2010 03:08:20 +0000

Thanx for the tip I was looking for this. Now lets see the results over the next week.

By: Velocity: The Ultimate Web Performance and Operations Conference | Phil Chen

Fri, 16 Jul 2010 16:45:18 +0000

[...] Kasindorf (dormando) and, after discussing file-system optimization, proceeded to implement some file-system tuning which, in turn, helped increase our Apache Web Servers running mod_disk_cache efficiency by 3 [...]

By: Vincenzo D'Amore

Vincenzo D'Amore — Wed, 09 Jun 2010 11:11:59 +0000

Hello,

I have mod_cache used in conjunction with mod_cband.
But when we enable mod_cache mod_cband are overridden.
Any idea in order to define the order of execution?

Best regards,
Vincenzo D’Amore

By: randy melder

randy melder — Wed, 05 May 2010 17:43:52 +0000

Nice job, Phil. Simple change with a big impact. I’m seeing lots of hits and first person accounts have proven this a valuable config change. You’re the man.

By: Phil Chen

Phil Chen — Wed, 27 Jan 2010 05:04:50 +0000

Doug in Paia,

Thanks for posting your results from your Apache ab tests. Looks like you are testing the theory on disk caching versus caching in RAM. Apache ab test is a good way to go about it however I think the variable is that of your disk speed as well as your EXT3 settings which maybe slowing your disk results up.

To be honest the performance difference between disk caching and mem caching in this instance is not very large my results regarding this aspect were different by just 0.2. The big factor is the EXT3 settings I outline when using disk caching that you get the bigger bang. I chose disk caching because it allows for a lot more cache storage versus that in the maximum amount of RAM in your server. So for my situation disk allowed me a lot more cache and the steps above gave me the performance I needed.

Thats just my view on it, hope that helps

By: Doug in Paia

Doug in Paia — Sat, 23 Jan 2010 23:13:15 +0000

Hi Phil, Thanks for the great post, this really got me jump started. I am running a small site with cakePHP. We host on a couple of CentOS 5 boxes, 64 bit on both VPS and dedicated. I am running apache ab against our VPS instances and I tried your settings. As well I tried a vanilla mem cache. I came up with the results below. The summary is that I didn’t get any performance boost for test on two sample pages. In the first instance, disk caching is slower by a factor of 3, memory caching is about the same. In the second test case, memory caching or not seems to be the same. Is apache ab a valid test? Any thoughts about what I might be doing wrong. Thanks, –Doug

n1000c10ModCache_new-releases.html:Requests per second:2.65
n1000c10ModMemCache_new-releases.html:Requests per second:8.59
n1000c10new-releases.html:Requests per second:8.89 << no cache enabled

n1000c10ModCache_moon-blue-ray.html:Requests per second:1.14
n1000c10ModCacheR1_moon-blue-ray.html:Requests per second:1.20
n1000c10ModCacheR1_moon-blue-ray_try2.html:Requests per second:1.23
n1000c10ModMemCache_moon-blue-ray.html:Requests per second:1.22
n1000c10moon-blue-ray.html:Requests per second:1.25 << no cache enabled

By: akira

akira — Wed, 25 Nov 2009 20:09:52 +0000

Did you try tmpfs as disk_cache?

By: Phil Chen

Phil Chen — Sun, 09 Aug 2009 23:37:30 +0000

Upon doing some research it looks like you are correct that noatime is a superset of nodiratime which is a subset. Thanks for the input Lamah.

By: Lamah

Lamah — Sun, 09 Aug 2009 11:20:29 +0000

Hi there, nodiratime is not actually required (it is implied as part of noatime, which applies to all files, including directories.)