Link to home
Start Free TrialLog in
Avatar of DavidBloom
DavidBloom

asked on

Virtualized Terminal Services performance

Would the user experience suffer if connecting to a Terminal Services server which itself was running as a Virtualized server?

The Hyper-V supervisor is Win2008x64 Enterprise.
The virtualized Terminal Services server is Win2008x32 Enterprise

Decent iron: Dual 5420 Xeon with 16Gb RAM

Thanks!
Avatar of tigermatt
tigermatt
Flag of United Kingdom of Great Britain and Northern Ireland image

If it's just one user accessing the server, then no - they would be pretty good with all that hardware and software to play with! Having lots of concurrent user sessions active on a virtualised Terminal Server still won't cause any issues, provided you don't go over the usual and recommended limits on CPU and RAM availability. Of course, in that sort of "double-virtualised" environment, continuous monitoring of resources allocated to each VM in Hyper-V is essential.

-tigermatt
Avatar of DavidBloom
DavidBloom

ASKER

Thanks tigermatt. Yes, common sense and experience would indicate that. But is it true?

Does a virtualized Terminal Server change published capacity expectations? Or for that matter, are there any findings that indicate reduced capacity BECAUSE it's implemented in a virtualized environment?
ASKER CERTIFIED SOLUTION
Avatar of tigermatt
tigermatt
Flag of United Kingdom of Great Britain and Northern Ireland image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Awesome. I'll introduce a virtualized RemoteApp server into limited production and see how it goes.

I'll be following up with optimization scenarios in another question. Hope you'll contribute ...

Thanks!
Thanks! Post a link here when you open your other question (to let me know about it) and I'll see what I can do.
David, I'm very interested in knowing how your research and testing went regarding the use of a virtualized terminal server.  I am *especially* interested to know how well it performs when under load.  I had very poor results using Microsoft's previous (non hypervisor) virtualization platform...  Even on very strong hardware, the CPU usage of the TS running under VS2005 was dramatically higher than one would expect after just a few users connected, and more than 5 people or so caused the TS to be virtually unusable.  I'd love to know i Hyper-V fares better.
Hi ccpd,

It's been 2 weeks since the "go-live" and I'm only just starting to catch my breath. Here's a summary of sorts recounting my experiences:

Taken strictly from a bandwidth point of view, running a 10-user medical management client\server environment from within a non-hypervised, fully virtualized single Win2008x32 server -- here it is -- PERFORMED MAGNIFICENTLY. The hardware used was not my biggest iron, but a test server based on an Asus P5E with a socket 775 version of the Xeon 5420 (the Q9300 quad-core) and 4Gb (the max) of 800ddr. So basically a phat workstation.

The server CPU, HD and NIC always had headroom under "normal" use. Looked like it would scale upwards to 100 users. There were potential saturation issues with Crystal Reports unless I forced all output query fields within reason, but that could also be resolved by performance "containerizing" the TS sessions with per-session performance percentage locking quotas.

Let me state unequivocally that the virtualized performance was ALMOST ALWAYS FASTER compared to the same client software running as a local install. Especially on the older P4 class systems. Only once in 2 weeks did I hear anything about keystroke lag, but that could have been attributed to a local process interfering with the TS session.

HOWEVER, I have since had to pull back from virtualization. Have done had to rebuild the server as a Win2003x32r2 because of some incompatibilities in the database manager running in the Win2008 (vista) environment. Thus, the next step (while the test server keeps the peace) is to tackle this production solution with a Hypervised Win2008x64 Enterprise Xeon server (ASUS DSBF-DE) running 4 virtualized servers:
1) a Win2003x32r2 runs the database manager (and perhaps other legacy software);
2) a Win2008x32 runs RemoteApp Terminal Services AND the virtualized medical management client software (and perhaps other vista-compliant software);
3) a Win2008x32 runs PDC/AD/DNS/DHCP and other "traditional" single-server services.
4) a Win2008x32 runs IIIS 7 services for TS remote gateway (to keep things secure).

What prevented my from keeping the virtualized server in place were:
+ The database manager not ready for Win2008;
+ TS Easy Print weirdness. Some reports would flat out refuse to print multiple copies. Other report formatting issues on network printers.
+ Need to lock down running only 1 instance of the virtualized software.
+ Nurses and checkout staff yegging about "too many passwords" and other timeout issues. They all hate HIPAA compliance. Too bad.

So ... now that Hyper-V is RTM, methinks I'll be diving into running virtualized apps from within a virtualized server. I have seen the virtualized future, and it (virtually) works.

Copacetic!

Comments?
David
I am about to enter into a very similar project and trying to hedge my bets. Were you able to build and test the Hypervised Win2008x64 Enterprise Xeon server (ASUS DSBF-DE) running 4 virtualized servers? If so how did it work out. Any other suggestions that will help me get off the ground?
Skip53