I have an issue that could well cause over 10,000 or 100,00 customers to participate. Has anyone done stress testing on a LimeSurvey when the machine performing the survey could be hit with thousands of inquiries an hour? I have not dug into the code, yet, but a very basic weak point comes to mind -- the database connections. Is there one static connection, or does each survey open a new connection? These kinds of things make me wonder whether LimeSurvey can scale. I'll find out the answer as I review the code, but I though I'm simultaneously post this thought on the forum and see what others might say.

And no, I'm not going to pay for the service of administering a survey as the survey would be done in a public interest fashion where the results are turned over to a government agency hoping to cause them to seriously consider the matter.

From what I have seen, one of the weakest point of Limesurvey regarding scaling is probably the size of the session files. However, there was once a sticky thread about Limesurvey with many respondents, but I can't find it at the moment. I guess it also depends a lot on what version you are using. I am not sure if there is a lot of experience with 2.5 with high load surveys, but 2.06 should be fairly stable regarding this.

Here a few additional tips: Don't send the emails out all at once, but send them in small waves over a few days, x per hour, etc. This will help to distribute the load over a few days and is generally best practice. Because the main problem is not the total amount, but the amount of concurrent access (how many respondents are in the questionnaire at the same time). It also depends on the size of the questionnaire and what machine you are running Limesurvey on.

I assume that you have a good email strategy, because with 10.000 or 100.000 email invitations sent, this can cause big problems in terms of spam if you don't have an impecable strategy (well configured smtp server, email text is avoiding any keywords that might be considered spammy, having a impecable list with double opt-in, etc.).

Thank you. Since posting this topic, I'm probably going to go for an anonymous survey, at first as visitors will be afraid of repercussions should they reveal who they are. So email management will not be a consideration. I guess I'll try scripting with Perl scripts to see if I can break the database connection. Again, I'm hoping there is just one connection to the database opened and that each session is not a separate connection.

I am not sure how Limesurvey handels the database.

Here a few threads about similar topics:
www.limesurvey.org/forum/can-i-do-this-w...ltaneous-connections
www.limesurvey.org/forum/can-i-do-this-w...vey-at-the-same-time

Maybe get in touch with Carsten, he should have some experience with scaling Limesurvey, as he runs Limeservice.com. But don't don't expect a quick response, he is quite busy. And today is a holiday in Germany. It is a pitty that the "scaling Limesurvey" thread disappeared. If I recal right, someone reported to have run a survey with 10.000 respondents in 24h or something.