Our non-profit organization has created a sizable public survey with over 450 questions and 1000 columns in the database (mysql). We have very complicated conditional logic which prevents most questions from being asked, so a survey should take only 20 minutes to run. We have done this without issue until we started getting a load on the survey (about 1 survey per minute). This has put a strain on our server, creating performance problems, and I suspect its due to the complexity of the survey (rather than bandwidth bottlenecks or slow php code).

I was wondering what other peoples experience with large, complex surveys and if they have had these issues and if they found ways of improving performance. Would it be possible to break the survey up into multiple tables? Would that improve performance enough to be worth it? What would be the best indicator that this is the problem?

Any help in this matter would be greatly appreciated. We have used PHPSurveyor/Limesurvey for a couple of years now with great success so far.

Hi, which version and build of LimeSurvey do you use?
On my knowledge on one of the latest build there was a speed improvement implemented by Jason Cleeland.

Hi,

I would take a look at your php.ini file too and make sure the settings are optimally configured.

You might also want to look at the following sites:

eAccelerator: http://eaccelerator.net/

xCache: http://trac.lighttpd.net/xcache/

Zend: http://www.zend.com/products/guard/optimizer/

I guess the problem is cuased by the condition checking. When having large surveys with lots of conditions there are lots of dependencies to check.

Please upgrade to the latest build which contains some speed improvings.

We are currently using version Version 1.71+ (5246), which is either the latest, or very close to that.

v1.71+ is the latest at revision 5448, so you're quite a way off the latest version from the repo. I think the speed increase came in at about 5397.

I will check out those sites. I did have to make changes to the php.ini file just to get it to run (i.e. increasing the limit of response variables).

Thanks.

I wondered if the number of conditions would be an issue, which is why I wondered if breaking the survey up into multiple tables would ultimately affect performance. The number of conditions is rather large (1,200 rows in the database).

I don't know the answer to that as I haven't run a survey with that complexity.

Could you break the survey into several and then combine the data post fieldwork?

Hi!
As said. First I would try the update to the latest version before trying anything else.

Second and quite easy step is eAccelerator: eaccelerator.net/

It is speeding up this website too.
If you should have still problems after that please write again.