Hi all
Has anybody tried answering the question "What is the size of the largest class my Moodle site can serve an on-line exam simultaneously?" through real simulation?
Basically, the same (standardized) quiz will be taken by a variable number of candidates during a variable time period - kind of ramping up - until the response time becomes unacceptable. Has any such thing been documented?
A starndardized quiz to benchmark a Moodle installation?
by Visvanath Ratnaweera -
Number of replies: 4
In reply to Visvanath Ratnaweera
Re: A starndardized quiz to benchmark a Moodle installation?
by Visvanath Ratnaweera -
Ping!
There is no such data set or is the question unclear?
There is no such data set or is the question unclear?
In reply to Visvanath Ratnaweera
Re: A starndardized quiz to benchmark a Moodle installation?
by Tim Hunt -
What is a standard quiz? Is your exam multiple choice quetions, Essays, or something advanced like STACK or CodeRunner. You are really much better load-testing with a quiz like your real one will be, and the work to make such a quiz will be trivial compared to doing the testing.
From reading these forums over the years, I have seen reports of people successfully running quizzes with hundreds of users at the same time.
When I was load-testing our quiz install, we simulated over 1000 users, but our live load never got that high.
In reply to Tim Hunt
Re: A starndardized quiz to benchmark a Moodle installation?
by Visvanath Ratnaweera -
Hi Tim
I feared that there is no such thing. Thanks for confirming.
To your questions:
> You are really much better load-testing with a quiz like your real one will be, and the work to make such a quiz will be trivial compared to doing the testing.
I have plenty of real quiz questions. That is not the problem. But for two people to compare the quiz performance of their Moodles, they need to take the same quiz! I was talking of a common/neutral quiz. I see various data sets coming from moodlenet, the Moodle MOOC, etc. waned to be sure, that i don't duplicate work.
> From reading these forums over the years, I have seen reports of people successfully running quizzes with hundreds of users at the same time.
I am not interested in getting the maximum number of quiz candidates at any (resources) cost, rather getting the maximum from modest resources. And to have quantitative results.
> When I was load-testing our quiz install, we simulated over 1000 users, but our live load never got that high.
I bet, you had big iron! As I said above, my idea is to compare smaller iron - quantitatively.
I feared that there is no such thing. Thanks for confirming.
To your questions:
> You are really much better load-testing with a quiz like your real one will be, and the work to make such a quiz will be trivial compared to doing the testing.
I have plenty of real quiz questions. That is not the problem. But for two people to compare the quiz performance of their Moodles, they need to take the same quiz! I was talking of a common/neutral quiz. I see various data sets coming from moodlenet, the Moodle MOOC, etc. waned to be sure, that i don't duplicate work.
> From reading these forums over the years, I have seen reports of people successfully running quizzes with hundreds of users at the same time.
I am not interested in getting the maximum number of quiz candidates at any (resources) cost, rather getting the maximum from modest resources. And to have quantitative results.
> When I was load-testing our quiz install, we simulated over 1000 users, but our live load never got that high.
I bet, you had big iron! As I said above, my idea is to compare smaller iron - quantitatively.
In reply to Visvanath Ratnaweera
Re: A starndardized quiz to benchmark a Moodle installation?
by Tim Hunt -
Actually, the system where I tested to 1000 simulated users was fairly moderate. 2 web servers and one database server. All VMs. I can't remember how many cores or RAM, but it was not excessive.
We did use Postgres database, which has always been good at simultaneous writes. I am not sure how good MySQL is in comparison. Probably better these days than in the past.
Anyway, if you get some numbers for what your hardware can do, I will be very interested to see them.