Hi mike,

>Recently I demonstrated that an exe on a server is not copied to the pc before it is run. In fact, screens and other components within the exe are extracted/accessed repeatedly while the application runs.

That observation is correct. You can check this with filemon (www.sysinternals.com). In general however the time wasted on network trafic to this exe might not be significant, even with a lot of users. However if the bandwith to the EXE is not wide enough, it definately is better to get this exe locally.

>In a citrix environment, I'm assuming that as each user launches an exe, there is a "virtual" machine started on a server. So suppose the main exe is on one server box. As each citrix virtual machine accesses that exe, there should be a performance loss.

Having the exe locally is going to be faster in almost every case. In my case the difference in speed was 50 about times. However there might be disadvantages to, because depending on if you've got a copy for every user or just one copy on the machine, a lot virtual memory can be wasted on (OS) caching the info, which could be better used for running processes. However I'm not an expert in tuning Citrix, so I can imagine there are mechanisms in place to efficiently deal with that problem.

Packing the exe with an exe packer like konxise and armadillo might have some other performance characteristics as the info in the exe is cached and unpacked in memory.

But all in all, I wonder if for the average installation it really matters. Having the right hardware platform might and a well balanced citrix environment probably is the best insurance for a good performance here.

>However, if the exe was copied to a location on the same physical machine that is running the virtual machine, there should be a performance improvement, right?

Personally I've found that most performance problems are related to slow commands and inefficient data access. Postioning of the executable does indeed matter over slower networks (10 mbps and below), but hardly make a difference on more recent technologies.

Again, if you're not already using filemon, you better use this tool to find your bottleneck in performance.

Walter,