>When you're "testing from a VM", I assume you mean you're testing from a TS session on the TS VM >against your VFP tables located in various places (?)

Yes, correct.

>Your point #1: I assume you mean you doubled the RAM allocated to the File Server VM (?) If you already were allocating 16GB to the TS VM, with only 24GB total you don't have enough to double the TS VM's RAM.

Doubled the TS from 8 to 16GB

>Under normal conditions, how heavily loaded (RAM and CPU usage) are both the File Server and TS VMs?
Neither seem too overloaded. TS uses more memory and CPU, as I expected.
>My first recommendation: if you're running antivirus with a real-time scanning component on either the File Server or TS VMs, disable it for testing.

We have set up exclusions to keep Endpoint Protection from scanning Fox-related files or network drives.

>Second, with virtualization it's easy to overload a disk subsystem. Your host's 15K SAS drives are fast for mechanical drives, but if there's a lot going on with the hosted VMs they could get overloaded. RAID5 is not ideal for write performance, either. In any case, to see if this is an issue, run PerfMon or its W2K8 equivalent on the TS VM while you're seeing performance problems. Check your results here: http://msdn.microsoft.com/en-us/library/ms175903.aspx

I still have to look into this. I’ll get back to you.

>If I read it correctly, your 3d test was "fast", using a TS session on the TS VM against VFP data files stored on a "local" drive on the TS VM (i.e. the TS VM's "C:" drive), rather than on the File Server VM. Also, all other tests where the VFP data source was the File Server VM were "slow". At first glance, it looks like either the File Server VM is slow, or the network connection to it is slow. It might actually be better to run PerfMon on the File Server VM during slowness and monitor its disk performance. If its disk performance seems OK, then it's a networking performance issue.

If I run the VFP app on the file server, with drive substitutions to create drive letters expected by the program, it flies. On either box, once the tables are considered ‘local’ the performance is a ten-fold improvement.

>Your 500MB file copy test was 4G bits and took 20 seconds, so performance was about 200Mbits/sec. We'd have to know if the disk was bottlenecked before concluding the network performance is sub-par.