XF 1.2 Measuring the load introduced by addons?

[Stuart Wright]

Is it technically possible to measure the amount of load added to the production of a page by each of the addons involved?
Clutching at straws here but if there are a few addons installed and one or more are increasing the load on your server, is it possible to measure the processing time of various chunks of the process?
It would be handy, though I have no idea whether it's possible.

 

[HWS]

Measure the standard load of your test-forum (without active real users) and disable one single add-on after another and measure again.

You'll get a least a clue which add-on is demanding more resources.

I like this: https://www.blitz.io/ for load testing.

 

[craigiri]

Measure the standard load of your test-forum (without active real users) and disable one single add-on after another and measure again.

You'll get a least a clue which add-on is demanding more resources.

I like this: https://www.blitz.io/ for load testing.

Not sure if that will tell you much since, IMHO, load under real stress is far different than load without....

If it were me, I'd probably start by narrowing down which add-ons probably do NOT add to load and then have a smaller group left to test - possibly on my live site, while watching the server load.

How about starting with some basic rules of thumb or comparisons?

We are currently serving about 2.2 million XF pages per month with a 3 year old server with a single quad core xeon and 8 G ram. Total DB is less than 5 GB.

This is being done without any stress on the server. I don't think we have ever hit a server load equal to the number of cores.

If you want a list of the add-on's I use, send me a PM.....this way you can eliminate some or all from your list as they never caused any noticeable delays for us.

I'm sure there are more technical ways to do this, but I like Rules Of Thumb.

 

[Mouth]

Activate debug mode to get the queries, load time, and memory in the bottom right corner.
Then activate/deactivated an add-on, refresh page add-on would have been actioning, and see the difference.

I've recently decided to drop bringing across a couple of add-on's in my 1.2 upgrade, using this method. I ruled them out because I didn't think the additional load was worth the value/features the add-on gave.

 

[craigiri]

Mouth, if any were especially CPU consuming, maybe mention them? I don't think that's a negative thing for authors because more forums are not really worried too much about these loads. I wouldn't be on my unpopular (40K pages per month) forum!

 

[Stuart Wright]

Problem with refreshing is that the page is cached after the first load.
We've been enabling a new one each day. But it's extremely variable.

 

[RobParker]

Measure the standard load of your test-forum (without active real users) and disable one single add-on after another and measure again.

That doesn't really work when the addon only causes problems with a lot of users online.

 

[HWS]

That doesn't really work when the addon only causes problems with a lot of users online.

This is why I recommended Blitz. It can simulate several thousands of users and measure how your website handles the load. You get a detailed report for each simulation. But you may not want "real" users to influence your load test.

If you start those simulations with different add-ons disabled you may get a hint what causes more load compared to other add-ons and how a single add-on may take down your server(s).