I have been doing some performance testing on Windows Server 2008, the goal was to see differences in how EPiServer CMS behaves on different operating systems.

These tests are performed using EPiServer CMS 5 SP1, the official support for Windows Server 2008 is coming in EPiServer CMS 5 R2 (I'll get back to some performance testing on R2 further on). The tool I used for testing was Visual Studio Team System 2008 Test Edition and the hardware was 2 dual quad-cores with 4GB of RAM, Ghost Solution Suite was used to setup the servers (I'll write a separate posting about our lab environment). I used 2 MS Load Test Agents.

Test 1

The first test I performed was to test raw template performance so there are no resources (images, JavaScript, CSS etc) downloaded, we are just hitting the templates. Its a 10% difference from the fastest Windows Server 2003 32-bit to the slowest Windows Server 2008 64-bit, not really what I expected but read on.

Test 2

In this second test we add output caching to the previous test. The fastest continues to be Windows Server 2003 32-bit and the slowest Windows Server 2008 64-bit, hmm..

Test 3

In the third test we remove output caching and add downloading of resources to the mix (images, JavaScript, CSS etc). This test is more real-world than the other tests, because who has a site without images? Now the roles change and Windows Server 2008, both 32-bit and 64-bit versions, are faster than any Windows Server 2003 version. Interesting, we are using the integrated pipeline if anyone wonders. But still, the 64-bit version are a few percent slower. Then if you are going to cache a lot of content (~1GB+) then you probably are better off with a 64-bit box and a lot of RAM (the few percent slower performance could turn around to better performance because you don't need to evict stuff from the cache).

What do you think, is this interesting ? Do you want us to publish more of these kind of tests ?