The small company I work for has two VMWare-hosted machines in the data center of a reputable, well established hosting provider. These VMs are running Windows Server 2008 R2 and are configured with Microsoft's RemoteApps software. There are about ten users logged in to RemoteApps at a time. Most of them are running the same relatively non-demanding Microsoft Access 2010 application.

While using the RemoteApp software, or logged in to Remote Desktop with an administrative account, there are intermittent (a few times an hour) pauses where the session will not respond to keyboard or mouse input. They are typically only a few seconds, but begin to grow both in frequency and in length (as long as three minutes or more!) as additional users log in. They occur on the RemoteApp server and the domain controller.

Our host has told us that this sounds like normal network latency, but increasingly frequent multi-minute pauses make me skeptical.

This peculiar latency is extremely frustrating. Has anyone encountered it before?

  • What have you done to profile this problem? Do you have procmon traces of the hanging process? Do you have performance logs that show what is happening on the server while these pauses are occurring? Is there anything relevant in the server's event logs? – MDMarra Feb 23 '12 at 22:24
  • Are the users all located in the same facility using the same internet connection? – xeon Feb 23 '12 at 22:26
  • @xeon No. The users are connecting from different geographical areas. – James Feb 23 '12 at 22:27
  • @MDMarra No; I didn't really know where to start. The event logs didn't appear to show anything abnormal that would affect this. I don't think we have performance logs, although I'm not familiar with Windows servers. I'm looking at Process Monitor now, waiting for the server to freeze up again. – James Feb 23 '12 at 22:45
  • Ah, I see. Well, hopefully procmon will shed some light. – MDMarra Feb 23 '12 at 22:46

It sounds like it could either be network problems (packet loss) or the server(s) that serves your vritual machines is (are) over utilized. Packet loss is easily detected, just ping the server for minutes, an hour or more and check the results. Normally you should have 0% packet loss, a couple of % of packet loss occasionally is not unacceptable. But if you have consistent packet loss something is wrong.

Run a traceroute to see if anything in the path may have problems.

If the server is over utilized, check my answer here: My server doesn't seem to be caching anything well

  • -1 ICMP really isn't a great tool to troubleshoot network problems and the answer that you linked to doesn't seem to apply here. – MDMarra Feb 23 '12 at 22:50
  • 3
    It is a great tool to quickly check if you have network problems, but not the only tool. Ping is an essential tool for any network engineer. I am rather dismayed you would dismiss ICMP so easily. The answer I linked to gives suggestions what to do when the VM server is over utilized. – aseq Feb 23 '12 at 22:55

So, I guess I didn't leave enough useful information here to start with, and I never updated the question with the event logs. But we did figure out the cause: processor usage. The CPU was pegged in the 90% range constantly, and when it hit 100%, everything halted until it had worked through enough instructions.

Making a second processor available to the VM improved the situation significantly.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.