Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations John Tel on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Terminal services: average bandwidth per session?

Status
Not open for further replies.

jollyreaper

Technical User
Jul 25, 2005
105
US
I've been trying to find some documentation on how much bandwidth (on average) each terminal services session should consume. As I understand it, if you aren't doing anything in the session and nothing is causing the screen to update, the consumption is just about zero. What is considered to be typical consumption? Exceptional consumption?

Thanks a bunch!
 
It's something in the range of 20k/session if I remember correctly. Exceptional consumption would be total screen refresh.
 
Thanks a bunch! I have another question about how terminal services works, assuming you have the answer. Remote tools like VNC seem to operate by sending screenshots of the remote computer you're logging into. Terminal services seems to be sending the rendering instructions instead, having the terminal services client then render the screen on your computer. Is this correct? It's hard to find documentation on this, at least with the searches I've used. Thanks again!
 
Hmm, this is odd.

I managed to find some Microsoft info on bandwidth consumption.


Network usage overhead tends to be quite low on Terminal Server, because of protocol efficiency and since, by default, the Terminal Services Client (mstsc.exe) uses data compression for all connections. Network usage for the two scenarios is shown in Figure 7. This includes all traffic coming in and going out of the terminal server for these scenarios.

(graph shows 1150 bytes/sec for "knowledge workers", 1290 bytes/sec for "data entry.")

Choosing 16-bit color depth for remote connection sessions instead of 8-bit slightly increases RAM usage (~10% when compared to 8-bit remote sessions), increases network bandwidth usage from 1150 bytes total/user to 1450 bytes total/user, but does not significantly affect the CPU usage level or the kernel virtual space consumption.


That sort of overhead strikes me as impossibly low. Then again, your 20k/sec struck me as depressingly high. Anybody have a rebuttal to this? Thanks!
 
I remember reading what you quoted a while ago in one of those TS white papers. I guess it's estimate... I usually use 20kb in my quick calculations... How many people will be connecting to your TS ?
 
Currently 7 to 10 remote users, will grow quite a bit in the coming year. 1.5kb is a nice, friendly figure. 20kb sets the heart a-fluttering. That tells me that I can fit maybe 9 or 10 active users. The 1.5kb figure is seeming closer to correct. I'm monitoring TS and see 7 online with active sessions, no idle time, and the overall bandwidth use is anywhere from 6% to 22%, depending on what else is going on. We're in the process of trying to install better monitoring software so we have more solid figures to work from.
 
Hmm.


An RDP Session can absolutely work with less than 56Kbps, 26.4Kbps is a better lower limit for RDP Sessions. It does not need 26.4Kbps just to type in notepad, but noone does just this, and minimizing/restoring apps, printing... will require at least 26.4Kbps.

Normal operation is pretty much the same as ICA with the same settings enabled. It's not that ICA is a much thinner protocol, but rather that the server side features allow better management of the available bandwidth. You can limit bandwidth of each virtual channel (via Citrix Policy) and of the overall session, enable image compression, local text echo...

Printing, animation, file transfer, larger display resolution and deeper color depth will all increase the required bandwidth of either protocol.

You can compare the bandwidth usage by looking at the counters in System Monitor Control -> Terminal Services Session -> Output Bytes.

Patrick Rouse
Microsoft MVP - Terminal Server


56kbits/sec, do we divide by 8 for bytes or by 10 to account for checksums? So something around 5k/sec with low-demand usage closer to 2k?
 
How remote are your users?

I have a TS set up with a 10/100 Ethernet, 2GB RAM, and 2x2GHz.
We get about 10-15 sessions LAN and VPN and the usage statistics make it feel as if nothing is happening. Processor usage is well under 5%, Memory usage is at about 1GB, Network capacity is under 1%.

All of the clients connect over VPN or LAN, using the application for PPT, Outlook and Adobe Acrobat/Distiller. 4 of the users are Outlook exclusive, and I use this TS server as my main administrative station, so 10 or so Mozilla windows, with 10 or so tabs each, Remote Desktops, Outlook, Terminal Services Manager, Exchange Manager, BES MGMT, Veritas Backup Exec, etc....

I think getting a beefy computer is the key.

I have used RDC over the new verizon wireless broadband cards connected to my VPN from a place with spotty reception. I set the RDC to treat it like a 56K session, worked great. There was some lag, I won't lie, but it was workable, with patience....


Robert Liebsch
Stone Yamashita Partners
 
Wow. That's pretty neat! Glad we won't be having to add a lot more pipe. There's a lot of growth planned and it will all be in the remote offices.
 
So, does anybody else have anything to add to this discussion?
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top