Creating a list of meaningful metrics to observe performance

Creating a list of meaningful metrics to observe performance

by Jerry McAllister -
Number of replies: 3

We are in the middle of evaluating Moodle to be used in a University setting with up to 50,000 registered students in a variety of courses and anticipating several thousand simultaneous connections. 

What we are hoping for here is to create a list of metrics that other Moodle sites have found to be meaningful in understanding what things are affecting the performance of Moodle at their sites.  Some things might be obvious such as memory use or disk I/Os.  Some may at first seem important, but you might have discovered them to be less significant.

We are not looking so much for a list of favorite tools to extract those metrics, though that could also be helpful, but a list of things to observe, especially to help anticipate possible problems.

Anyway, any observations you can contribute to such a list will be helpful.  I will try to keep a tally and summary of comments and put it up either here or somewhere easily accessible.

Thank you for any information you can give.

////jerry  

Average of ratings: -
In reply to Jerry McAllister

Re: Creating a list of meaningful metrics to observe performance

by Tim Hunt -
Picture of Core developers Picture of Documentation writers Picture of Particularly helpful Moodlers Picture of Peer reviewers Picture of Plugin developers

http://learn.open.ac.uk/mod/oublog/viewpost.php?post=65778 has screen-grabs of some of the OU monitoring tools. http://podcast.open.ac.uk/oulife/podcast-OU-moodle may also have some useful infomation.

We use JMeter for load-testing.

In reply to Jerry McAllister

Re: Creating a list of meaningful metrics to observe performance

by Jerry McAllister -

Thanks for the responses so far.  I am aware of jmeter, but it looks to me more like a load stress predictor and we are now looking more at monitoring currently running live systems.

Anyway, what I am hoping is to create a list of things to monitor more than the tools to do the monitoring.   The question is which things are actually meaningful to monitor on a live functinoing system that will help give a picture of how close we are coming to problems -- on those live systems. 

So, as an example, would charting memory use be significantly useful?  How about database hits - reads and writes of new records.  I am sure those both are valuable, but what are some others?

Thanks,

////jerry