• 4
name Punditsdkoslkdosdkoskdo

Terminal server performance over high latency links

Our datacenter and head office is currently in Brisbane, Australia, and we have a branch office in the UK. We have a private WAN with a 768k link to our UK office and the latency is at about 350ms.

The terminal server performance is reeeeealy bad.

Applications that don't have too much animation or any images seem to be okay. But as soon as they do, the session is almost unusable. Powerpoint and internet explorer are good examples of apps that make it run slow. And if there is an image in your email signature, outlook will hang for about 10 seconds each time a new line is inserted, while the image gets moved down a few pixels.

We are currently running server 2003. I have tried Server 2008 R2 RDS, and also a third party solution called Blaze by a company called Ericom, but it is still not too much better.

We currently have a 5 levels dynamic class of service with the priority in the following order.

  1. VoIP
  2. Video
  3. Terminal Services
  4. Printing
  5. Everything else

When testing the terminal server performance, the link was monitored using net-flows, and we have plenty of bandwidth available, so I believe that it is a latency issue rather than bandwidth.

Is there anything that can be done to improve performance. Would citrix help at all?

      • 2
    • This may or may not be relevant to you but I have noticed that any traffic I've sent through the Brisbane Juniper network (owned by Optus) always seems to have excessively high latency.

Yep Citrix is the way you probably want to head with this. RDP's vanilla implementations are great for low-cost solutions under most high-bandwidth, low-latency environments for general desktop/windows app distribution. But the whole thing kinda falls apart under specialist workloads and high latencies.

I answered a recent, relevant question where I think there's some crossover, here: Improving Performance of RDP Over LAN

You'll definitely want to test thoroughly before splurging any cash, though, as you still may hit limitations with such a high ping.

Outside of the Citrix/ThinApp solution, you may need to consider decentralising some of your applications and moving them back towards the branch offices. Even if you can't move them right out to the branch, having a rack or some kind of presence in a datacenter in the UK or europe, and hosting your solution out of it, may be the best option.

  • 3
Reply Report
      • 2
    • Have another +1 for Citrix, although I suspect that in this case even that may not be enough. 350ms is an awful lot to try and overcome.
    • +1 Citrix ICA is definitely more optimized for low-bandwidth, high-latency situations than RDP and can also be tweaked much further - but there's no magic to it, some things are simply not possible ^^

Either increase the speed of light, or accept that a third of a second latency is going to be very perceptible to humans and replicate some systems in the uk office.

  • 0
Reply Report
      • 1
    • Worked this out a few weeks ago: Speed of light = 299,792,458 m/s Circumference of earth = 40,075,020m. So theoretical minimum lag for a full trip all the way round = 134ms. Everything else is switching latency, modulation etc.

I think your best bet would be to stop using TS to Aus. for most applications in the branch office. I don't see any reason why you could not host applications in the UK, which would be a terrific improvement, latency-wise. That might save you the cost of buying Citrix licenses.

  • 0
Reply Report

Warm tip !!!

This article is reproduced from Stack Exchange / Stack Overflow, please click

Trending Tags