Hi,

We have caught a tiger by the tail. One of our surveys is going to be publicised on UK BBC TV and radio in April.

I have no idea how we work out what the potential load issues might be like. We might potentially be talking about 1000s of concurrent potential surveys simultaneously. Then again it might be a complete flop and we only get a few hundred.

Are there any software bottelnecks within the LimeSurvey structure?

Do you have any experience of mass use of the survey?

How robust is LimeSurvey against MySQL injection or other standard hacker tools?

Any help or suggestions would be greatly appreciated.

Kind regards
Chris

55chris55 wrote: Great!

55chris55 wrote: Possible bottlenecks might be:
- having lots of (huge) questions and displaying all on one page
- lots of complex and chained conditions
- huge images or the like
55chris55 wrote: - Some users have collected 20.000 responses a day so this should not be a problem if the server hardware (memory, CPU speed, several CPU cores) provides the necessary ressoureces. You won't be able to deal with that on a shared hosting.
55chris55 wrote: Yes, it is tested regularly and there are no known issues
55chris55 wrote: To my knowledge our hosting service was improved recently to use a server grid when necessary if you need lots of ressources. Have a look at www.limeservice.com and ask Carsten (c_schmitz) about the details.

Following up on this old thread, I recommend that the LimeSurvey team spend part of the 2012 Fund-Raiser funds to create formal load-test metrics for 1.92 and 2.0 for several representative surveys so that users can properly assess the sizing they need to launch their surveys.

I have worked with several load testing tools in the past, but all require a client install and can only generated limited load unless you have a cluster of machines.

I see that there are now commercial, cloud-based load testing sites (such as LoadImpact - www.loadimpact.com) which can generate true load testing from around the world for relatively minimal cost (you could provide reasonable benchmarks to the LimeSurvey community for about one to two thousand dollars US).

As to which sites to load test, I recommend provisioning a dynamically scalable cloud-based platform (such as those offered by 1&1) so that you can test several standard configurations (e.g. small, medium, large cores/RAM) cope with small/medium/large user loads. And of course, you'd also want to formally load-test the LimeService platform too.

Hi Tom,

I also consider this information very valuable. Please get in touch with Carsten about it directly. He doesn't check the forums very often.