CiviCRM Community Forums (archive)

Hi Dave,

I'm going through the same sort of process as we speak and a bit of learning as I go.  My contact database seems to be in the same order of magnitude as yours although it is on Drupal.

Mine is on Drupal though so the times you report don't really mean all that much to me.  How fast are the developer tools in Chrome or Firebug in Firefox reporting that the pages load?  I am starting to use Apache Benchmark too to get some numbers, but I've not sorted out how to test anything useful as it requires a logon.  From the browser I'm getting about 2s page load times which is roughly what the demo site shows.  Users are based in Australia and the host is in the US (Hostgator) so contemplating moving the host to an AU hosting based on tracert reports.

I found the single biggest performance improvement was enabling mod_deflate on my apache server, but have seen some improvements with tuning mysql as well, but it's all a bit hit and miss for me at the moment.

Malks.

Hi,

Interesting discussion.

First is to have real mesures. pagespeed yslow and the net panel on firebug allow to have real data (eg it took 7.48 seconds for page X, way more useful than "it's kinda slow").

In my experience (we spent a lot of time tuning): have a php accelerator, group js & css (on drupal), don't load a lot of modules, avoid acl, compress what you can, be sure to add an expire on all what's static, tune mysql, use lighttpd/ningx instead of apache (or as a front-end).


Otherwise, the biggest issue for speed is the number of requests. Some features (eg dashboard, editing a contact) do way more ajax requests than needed, and it's slowing a lot civicrm (we know how to fix but don't have the time/money).

And benchmark. Compare with demo.civicrm.org (sane configuration, without going all crazy on the tuning) from the same browser for the same task. If it's a difference for all pages, likely that's a network/server configuration issue. If it's only for some pages, might be that civicrm needs some mysql tuning (so it works as fast on 4k than on 102 contacts). Or at least, you know where to look first and what's likely to make a big difference.


X+

Hi,

Thanks for all the replies - I need to respond more fully when I have more time.

I am already using firebug to try and get some more hard and fast stats thanks for the pointer to 'net panel' this was switched off and I was looking for this very thing. I compared Joomla stats and they are similar in both locations. I'll try and post some net panel stats.

Certainly will compare to the demo site - dope - didn't think of that one.

Hershel - would be interesting to see your independant reports showing underperformance? Can you PM me?

I'm with Rochen who seem to be well regarded for Joomla hosting - I think that is what I meant by premium hosting.

My comment about normal was to try to get people to share their experiences to do just as you commented to compare speeds across different machines.

Could we set a few benchmark processes to do (open dashboard, edit contact, run search x) and a standard monitoring tool (net panel) and then share results so we can build up what might be normal?

The general comment about the MVS was that I am on a shared host where I have use of up to 8 CPUs (fair use of 2% CPU time in 24hrs) and plenty of RAM. I know the server isn't battered by watching WHM server load. They said that if I moved to an MVS (there are lots of benefits obviously in terms of security, isolation, dedicated resources) but I would only get dedicated 1 CPU equivalent. Thererfore not necessarily a performance boost. Any comments?

I was in your understanding about better performance on an MVS but I can also understand what the guy is saying. If the shared server was running at 7 CPU constantly then my CPU usage may be restricted but seeing as it running down at the 2 CPU level there is plenty of head room for civi to use more CPU if needed and they can't see how they can activily throttle because there usage policy is on a 24 hr period.

Your thoughts would be appreciated.

Cheers,

Dave

PS - potetially dumb question coming - if the 2 locations are getting similar network bandwidth and serving the same pages from the same server with different speed results then can there be sometihng to do with the local machine spec? I thought that seeing as this runs in a browser this wouldn't be the limiting factor but does ajax and other scripts take up local processing performance, IE if my machine is grunty and theirs is older will this have a noticable impact?