Terminal Server 2003 via VPN - poor performance; why?

From: Dr Him <geek_at_heurikon.com>
Date: 28 Oct 2004 13:09:55 -0700
Message-ID: <f47f1faa.0410281209.64df6fb6_at_posting.google.com>

We've got two sites, our main site in Madison, WI, and another in Edinburgh, Scotland. They are linked together with a site-to-site VPN. The Internet connection is a full T1 (1.5mbit) in Madison and an E1
(2mbit) in Edinburgh.

We've got an Oracle Applications (11i) server running in Madison. Because the connectivity between the sites is limited, we installed a Windows 2003 Terminal Server in Madison to provide a faster user experience for Edinburgh users.

Initially it seemed fine. However, upon actually having to use Oracle Apps across the pond, the users started complaining of horrible performance. Some times of the day it is fine (towards the end of the workday in Edinburgh), but much of the time it is pretty abysmal, taking a very long time even click a button (as for UI feedback, it's got to redraw to show the button being clicked).

Looking at the latency, it's pretty steady throughout the day at just under 200ms, which isn't great, but it's not *that* bad:

http://johnstewart.com/ACP/edinbdc-3h.png
http://johnstewart.com/ACP/edinbdc-30h.png
http://johnstewart.com/ACP/edinbdc-10d.png

And figuring throughput would be the problem, I used TTCP
(http://www.pcausa.com/Utilities/ttcpdown1.htm) to try and gather
throughput stats. I did a throughput test every 5 minutes, and it generally is pretty good, at 35-40 kBps, which is 280-320kbps, still pretty awesome compared to a modem (upon which Terminal Server should work relatively well):

http://johnstewart.com/ACP/throughput.jpg

I personally use the terminal server all the time, and it always works fine for me in any application. I do normally use it via decent links
(DSL, usually), however. But there are rarely any users on it, so load
is not an issue at all as far as resources on the TS box itself.

I'm not seeing anything obvious pointing me towards a possible problem.

As far as tuning the actual application, our Oracle folks are saying that there is not much that can be done to the client (to stop unnecessary redraws, for example).

Shouldn't our bandwidth and latency levels be adequate?

Does anyone have any suggestions on how to track down the problem?

thanks

johnS Received on Thu Oct 28 2004 - 22:09:55 CEST