Windows 200x impressions after using it for testing

We are in the midst of Solaris 10 testing for a customer. Explaining why Linux is so much faster (and more stable) on the hardware is getting old. So I’ll take a break and talk about the windows 200x experiences we had recently. A customer wanted to see performance on a number of things running on Windows 2003. They had a particular application that runs on it, and wanted to see what we could do with JackRabbit. Sure, no problem, we like doing these sorts of tests.

So we took two drives, loaded one with Windows 2003 32 bit Enterprise Edition R2, and the other with Windows 2008 server 32 bit enterprise edition. Customer wanted 32 bit. Well, not correct. The customer wanted whatever the app wanted. The app wants 32 bit. Ok. Took them both through setup/installation, driver loads, tweaking/tuning. I am not pleased with windows 2003, it has some rather glaring performance limitations. Doesn’t deal with high network or IO loads well at all. Loss of performance relative to Linux is pretty significant at the high end of workloads. It didn’t feel/act very responsive. Installation was actually a challenge, as I had to change bios settings to remove AHCI from the disk config for the boot drive (or it would BSOD on startup). Windows 2008 is another story. Completely different story. First off, installation did not suck. I cannot emphasize enough that, for the previous history of all windows I have ever installed, installation was somewhat of a crapshoot. It would sometimes work. I have taken to a minimalistic installation approach given our problems in the past … get a working version going first, install useful stuff second. But more importantly, unlike W2k3, the network adapters didn’t take their settings as advisory. Setting the IP addresses actually stuck (unlike W2k3). Second, usage did not suck. The interface took a little getting used to, but it was fast/responsive, didn’t crash. The performance monitoring tools were OK, with the caveat of a few (cough cough) humorous bugs (see the negative IO bandwidth image). Apart from its sense of humor, it was quite easy to do what I needed it to do. Granted, out of the box, IE is useless. Installing Mozilla/Firefox on there is easy and strongly recommended, as it is both more secure and less annoying. Third, and quite important … performance didn’t suck. This was the closest to Linux performance I have seen. I could push it hard, and it responded well. IIS7 didn’t suck as bad as IIS6 did (we had to redo some of the streaming tests on W2k3 with Apache just to be able to handle the load … IIS6 fell over under the load). But using more than one NIC on a multiport NIC card didn’t have a severe performance drop off (as it did in W2k3). Fourth, integration into the environment was quite good. Things that should have just worked, did just work. I would have liked Cygwin to be included … I tried SUA, and it is not as good from the compatibility standpoint as Cygwin. Cygwin could be adapted to use SUA backend for best performance, and best compatibility, and we would recommend this to Microsoft. All in all, from performance to usage, to integration, Windows 2008 left a distinctly different impression on me than I had anticipated. Its good. We would recommend it and use it. Yeah, thats right. (we are not biased towards/against vendors, just biased against crappy products, misleading marketing/benchmarking, and tools). If we can work well with something, we assume our customers can as well. If we cannot, we assume it will be hard for our customers as well. Now we are going to pull down the HPC server 2008 and see how that goes. Good job on 2008 Microsoft.