The long awaited NAV 2009 Performance Guide is here. You can download it from the mbs.microsot.com site.
You should download the document and read it to draw your own conclusions from the tests that were performed. I found the forty-page document quite heavy going with a lot of technical results (that I didn't fully understand) and some limited conclusions of the analysis. I did find the summaries useful, it's just that I could have used more summary and less data.
One interesting conclusion, for me, was that a single NAV Server appears to be able to comfortably handle 50 concurrent users. This was based upon the NAV Server application, being a 32-bit application that can work with 3GB of RAM or less, showing a 1.7GB memory consumption for 50 concurrent users. Obviously this is dependent on the type of hardware you are using and the scenarios your users are likely to be running. The document contains sufficient disclaimers on the findings so, as is often the nature of benchmarking results, this information is more of a guideline than a rule.
With respect to virtualisation, the team do not recommend using a Hyper-V virtualised environment for production use due to the reduced performance.
Personally I found that this document fell short of my expectations. It is clear that the team have put a lot of effort into their studies; however, I would like to see a guide to sizing hardware that is easy to use and understand. It would be nice if we could have some recommendations that answer the following questions:
- What is the maximum number of users?
- At what point should I separate the tiers on to three machines over two?
- How many users should I be running per NAV Server?
I would also like to see the scenarios run in the Classic architecture to see what performance benefits we get from the three-tier architecture.