Testing server load capacity

[Naughty]

Is there a tool to test a new server for load capacity?

We're moving servers, but before making the final switch, we'd like to 'hammer' it with some requests to see it it can cope. I know i saw tools before, but i forgot where.


You know of, or have such tool(s)?

[BladeZ]

google.nl/search?rlz=1C1GGGE_nlNL460NL460&gcx=c&sourceid=chr ome&ie=UTF-8&q=new+server+load+test

[CurrentlySober]

Huhh.. Huhhhh... You said 'LOAD'...

[Naughty]

google.nl/search?rlz=1C1GGGE_nlNL460NL460&gcx=c&sourceid=chr ome&ie=UTF-8&q=new+server+load+test

Already there obviously. Rather have recommendations than salespitches.

lol @sobermofo

[Jensen]

Visual Studio? It includes a load test tool which enables a developer to execute a variety of tests to simulate real user load.

Others could be Apache JMeter (desktop client) perhaps?

[CurrentlySober]

Huhh.. Huhhhh... You said 'LOAD'...

& "Tool"...

[marlboroack]

Huhh.. Huhhhh... You said 'LOAD'...

meehee huh

[dubsix]

http://en.wikipedia.org/wiki/Web_server_benchmarking

[AdultEUhost]

What OS are you running?

I would ask your host to hand you over some performance test results, if your box is managed that shouldn't be a weird question in my opinion and that will give you an idea about if they know what they are doing or not too...:2cents

[raymor]

ab (ApacheBench) is the standard tool.
http://httpd.apache.org/docs/2.0/programs/ab.html

In practical terms, the CPU and RAM are about a thousand times as fast as the drives, so one a web server the capability of the machine is limited by the drives. Therefore, drive specs tell 90% of the story. That and check whether the machine is using the "noatime" setting. Noatime will increase performance by 15%-50%.

That's the capability of the machine. What about software? Well, the server software pretty much just sends a message to the OS asking it to read a file from the drive and send it to the network card. Assuming a gigabit network card, the drive is again the slow part. The web server software really doesn't make much difference as long as it's not doing something extremely dumb. The main thing it can do or not do is force noatime despite the setting. lighttpd, for example, accidentally forces noatime even if the server admin has said not to use that setting.

So anyway, for most sites, check the drive and RAID controller specs. That's basically your capacity right there.

[CYF]

ab (ApacheBench) is the standard tool.
http://httpd.apache.org/docs/2.0/programs/ab.html

+1 for ab, that's what I was going to suggest. Nice tool for testing webservers.

[raymor]

Having said what I did Bonnie++ might be a good year to run as well. I'm assuming a pretty typical porn site load. A massive account of traffic to a poorly written PHP script such running say FHGs on a poorly designed database is a different type of load that will suffer more on different hardware. A lot of video encoding, say re-encoding each time a video is played, is something else again. Your test needs to represent the type of load you have.