Our current production SQL Server environment is running on AMD Opteron processors in a quad socket host. Due to licensing shifts and the track record of improvements in single threaded performance in a 2 socket Intel server over the quad socket AMD we are trying to conduct evaluation of the proposed new infrastructure. Our strategy is loosely defined due to our lack of experience and tools to accomplish this comparison. We have been leveraging Distributed Replay along with performance captures from our production environment for the same period to facilitate a close to real-world execution profile. However, Distributed Replay isn't all it is marketed to be when running it across multiple clients. As a result I am being asked to determine the best approach to conduct this testing and what I am interested in knowing is if any of you have also made similiar transitions in infrastructure for production SQL Server databases and how you went about evaluating the execution performance. Any tools or utilities used would be helpful. The storage infrastructure, network infrastructure, and general I/O statistics from artificial load have shown that the new infrastructure is at or above the same performance level in each of those categories. However, without comparison with a real-world simulation of the OLTP workload across the two servers buy in would be questionable.
Any help you all could give with some direction would be much appreciated.