Web Hosting Talk : Web Hosting Main Forums : Web Hosting : Website load testing

[Mize]

I have been developing a few sites for some time now, and everything is starting to come together. I decided to do some load testing to see try and estimate how many concurrent users I'm going to be able to handle without causing undue stress on the currently shared server.

So I downloaded a stress tester that lets you record yourself browsing and replays the clicks/visits it X number of times that you specify. But before proceeding I was wondering if there is any particular etiquette I should follow in load testing a site on a shared server. I figured I would do the testing at non-peak times, and wait until the server load is someone low, just to avoid slowing any one down. But beyond that I'm not sure currently how to proceed.

p.s. I am aware load testing isn't 100% realistic, but right now its the best I can do other than taking the site live and hoping for the best.

[CodyRo]

I think a few hundred requests at a time would be fine - do whatever you think would be reasonable with a shared provider (or in other words: use your own discretion).

Not much help I know, but if you're trying to emulate real users / realistic traffic it shouldn't be an issue.

[040Hosting]

Originally Posted by Mize

I have been developing a few sites for some time now, and everything is starting to come together. I decided to do some load testing to see try and estimate how many concurrent users I'm going to be able to handle without causing undue stress on the currently shared server.

So I downloaded a stress tester that lets you record yourself browsing and replays the clicks/visits it X number of times that you specify. But before proceeding I was wondering if there is any particular etiquette I should follow in load testing a site on a shared server. I figured I would do the testing at non-peak times, and wait until the server load is someone low, just to avoid slowing any one down. But beyond that I'm not sure currently how to proceed.

p.s. I am aware load testing isn't 100% realistic, but right now its the best I can do other than taking the site live and hoping for the best.

In my opinion you should contact you host; I can however tell you what I would respond in such a case;

Share hosting is not a development platform, you should not perform any kind of tests on it, for this you better take a VPS, which is many ways is better to do these type of tests.

But your host may think differently and let you do such a test. I believe that in most cases your IP simply gets blocked by the firewall before it can do any damage.

[MikeDVB]

As long as you don't do anything outrageous (like bring the server down) or increase the load by more than just a little bit you will be fine and likely the host will not even notice.

If you crash the server or you jump the load up by a good amount be prepared to at the least get an email, or maybe the boot.

[Mize]

Well, I don't intend to bring any servers down
My test will consist of 10-20 concurrent users on the website. I'll start with 10 and see how the resource usage goes, if its low enough ill try more.

About the vps recommendation... I've found that in a vps environment, page load times/load production aren't always consistent. IO waits and such caused by other users effect my load times more than script tweaking. I guess the same could be said for overloaded shared.. I guess I could rent a dedicated server just because "shared hosting is not a development platform" but that doesnt really seem necessary to me. Particularly when the test simulates real browsing...

Went ahead and did the test, everything went pretty smooth. <5% cpu at first, and dropped to 1% after the pages were cached.. not too bad.

Thanks to everyone for the advice

[040Hosting]

Originally Posted by Mize

Well, I don't intend to bring any servers down
My test will consist of 10-20 concurrent users on the website. I'll start with 10 and see how the resource usage goes, if its low enough ill try more.

About the vps recommendation... I've found that in a vps environment, page load times/load production aren't always consistent. IO waits and such caused by other users effect my load times more than script tweaking. I guess the same could be said for overloaded shared.. I guess I could rent a dedicated server just because "shared hosting is not a development platform" but that doesnt really seem necessary to me. Particurally when the test simulates real browsing...

That is maybe because i call 10 to 20 concurrent users not a stress-test. make sure you inform your webhost about that amount if you would contact them at all.

[Mize]

Originally Posted by 040Hosting

That is maybe because i call 10 to 20 concurrent users not a stress-test. make sure you inform your webhost about that amount if you would contact them at all.

Well, maybe stress test isn't the best description, really i'm just looking to see if my site needs further tweaking. Don't want to go live and find out a moderate amount of visitors is going to peg the cpu due to some script freaking out
But after the test, things seemed to hold together.

[040Hosting]

Good to hear it worked out for you, as you mentioned before; it is depending on many factors; and yes IO and VPSes may cause some issues, but as you said that is not much different as from a shared-host.