Slow Remote Terminal Service Sessions 1

[snootalope]

Ok, I know this has probably been discussed a million and one times, but here's another one!

I've got remote users that access our terminal servers from home. They're all on broadband connections.

Our company's internet backbone is a full T1, our terminal service office is connected to our ISP via two bonded T1's. We're barely using a full T at the moment (just WAN traffic, not internet traffic, it's not even close to using a full T1).

I've dropped the color depth to 8bit, disabled all the extra junk like the Windows Effects etc.

They still complain about sometimes while typing their text will take sometimes minutes to appear on the screen.

Our hardware:

Home user -----> Internet ----> PIX 515 Firewall ----> Cisco 1751v router ----> Netgear SSL312 SSL/VPN provides clientless remote access to our terminal servers

The Netgear is fully updated as well.

The terminal servers are pretty well built machines. Dual 3.2ghz, 3GB of RAM, 10k SCSI drives, and at the heaviest load they'll have about 24 users on each.

The Netgear uses a built in RDP to connect to the TS's.

So, what in the heck. I've been fighting this for YEARS and it never seems to go away. Honestly, there's got to be something better.

I'm considering dumping the Netgear and upgrading our PIX to feature the WebVPN/Access Anywhere, but I don't if that's any better than what we got, plus our remote access MUST remain strictly clientless.

Any advice? Ideas? Best Wishes?

 

[Potomac]

I'm guessing the problem is more an issue of latency than bandwidth. Terminal services with RDP clients needs very little bandwidth, but high latency times will kill your performance.

 

[snootalope]

From where to where should I test the latency and how can I do that?

Like from the client would I just use ping or something? Still, how could I determine the full client-to-terminal server latency?

 

[Potomac]

I'd start with ping, from one of the remote sites that complains of speed problems. Make sure your different devices are set to respond to ping, and hit each one from your remote site. I wouldn't be surprised to see the latency jump somewhere between the pix and the final server.

 

[snootalope]

so what would you consider a bad round trip time? From our PIX to the final server(which crosses the bonded T1's and the router) averages probably around 15ms during a normal load time. Not bad at all. I'll have to check what it is on the client to pix side.