My employer has over a thousand servers (running SQL Server 2005 x64 and a couple of other apps) all across the country. And in my opinion they are all massively underpowered for what they need to do.

Specifically, I feel that the servers simply do not have enough RAM for the amount of volume the machines are asked to do. All the servers currently have 6GB of RAM. The users are pretty much always complaining about performance (mostly because, immo, the server dips into the paging file quite often).

I finally convinced the powers that be to at least try out a memory upgrade on one box and see the results. However, they want before and after metrics, so that they can see that the expense will be justified.

My question is what metrics should I collect to see whether the performance truly improves on the box? I am a dev, so I am not sure how and what to collect (i have a passing knowledge of Perfmon).