Windows 2003 Terminal Server slowdown issues.

CruisD64

2[H]4U
Joined
Mar 6, 2007
Messages
2,464
Hey everyone. Merry Christmas Eve!

I work for a construction company. At my workplace, we rely heavily on Terminal Services for field jobs. At any given time, we may have up to 35 people on the system. The system specs are as such (I think they are more than plenty):

Dual Xeon E5345 Processor's (2x4x2.33ghz)
14gigs of ECC DDR2
Windows Server 2003 R2
80gig primary drive. Network drives totalling over 4tb for storage.
100mb Ethernet connection

The network connection is a bonded T1 (3 up / 3 down) line. I know this is not enough bandwidth for optimal usage, but for now it's our only option and seems like it should be enough. T3 would be great, but cost is an issue and let's face it, we're in a recession (so tired of that word).

I have a few questions regarding things that we could do to speed things up. For now, people are connecting to Terminal Server using the FullScreen setting as well as 32bit color. Is there a way to manually force lets say 256 colors or 4bit colors as well as a static resolution like 1280x1024?

Also, would there be any benefit to upgrading to Windows Server 2008? I know that it has an up/down managing feature where you can manually set the priority for sending and receiving data. It seems like a neat function, but would it help towards my end goal?

The programs that are most commonly used within TS are FileMaker Pro to access our old database which we are currently phasing out, and Viewpoint Construction Software which we are transitioning to. It seems we have plenty of ram because the most I have ever seen used up is maybe 8 out of the 14 gigs.

Also, if there are any other tips or tricks that you may have which would help speed things up, please let me know. I'm not the primary Sys Admin at the company, so I can't make any decisions without consulting my supers, but I would like to bring something to the table so that we can get things moving.

Thanks for all your help!!!
 
Do you know for a fact that the T1s are the bottleneck and that RDP is a protocol that is consuming a significant portion of the available bandwidth?

Think about it like this: If you have 3x1.544 Mbps available minus overhead and RDP was only using 10% of that, reducing RDP's usage by say 10 or 20% is not going to really speed anything up noticeably.

If you want to go ahead and force clients that are connecting to the server via RDP to have specific settings enforced you can do that using Group Policy. Windows 2003 has fewer configurable options than 2008 though.
 
When I see "80 gig drive"..makes me think SATA drives, yes?
What antivirus is on it?

I take it applications are run from the network drives....is it going through a gig switch? (N/M...just backpeddled and saw your 100meg connection). Any gigabit ports on the switch to try?
 
I don't know for sure that the bandwidth is being fully saturated. I've requested bandwidth statistics from telepacific in the past before and they have shown full saturation. I will be getting updated statistics shortly.

As for the applications, they are being run off of the local drive with data being stored on network drives.

Am I unable to force lower image settings through the host computer?
 
Before you think about solutions it would be more efficient to identify the cause of the saturation.

If you know the T1s are max'd out then identify which protocol is the culprit, i.e. RDP, HTTP, FTP, torrents, etc. My current customer has just as many users on a single T1 which is used for remote administration of AIX, Linux, and Windows machines via a VPN and it's never close to max'd out. Once you figure out which protocol it is you can better identify who the "hogs" are.

Back your original question though, if you have access to the server in question you can examine the settings available in group policy by running gpedit.msc and drilling down to the terminal server settings. Server 2003 doesn't have as many options available as 2008 (which I haven't played with since I haven't been a sysadmin for ages) but it should still have some settings in there that can affect bandwidth.
 
I don't know for sure that the bandwidth is being fully saturated. I've requested bandwidth statistics from telepacific in the past before and they have shown full saturation.

I'm just curious but why can't you find out your bandwidth statistics from your side?
 
When I see "80 gig drive"..makes me think SATA drives, yes?
What antivirus is on it?

I take it applications are run from the network drives....is it going through a gig switch? (N/M...just backpeddled and saw your 100meg connection). Any gigabit ports on the switch to try?

I see that 80gig drive and facepalm. With a handful of users, sure...

What would be the best utility to check realtime bandwidth statistics?

Taskmanager>networking. Do this on the file server where the db is and on the server where people are coming in from the outside.

My thoughts are an IO bottleneck. Try a faster disk.
 
But the thing is. When using TS from inside the network, it runs great. Even when people are reporting slowness from the outside. That's what makes me think it's a bandwidth issue. Granted, I know you can't expect LAN speeds when accessing remotely. I will get stats this next week and update this thread.

Happy Holidays everyone!
 
maybe it is a bad router that cant handle the sessions?
 
Check out Doug Brown (http://www.dabcc.com/) He had an excellent article on speeding up Terminal services. Now before you dismiss this because of citrix, do check it out. A published desktop session really is no more that rdp in my book.

I do have to agree with everyone else that the t1s are more than likely your issue but you might as well cover all angles. Also it would hurt to invest in some gig switching for your servers

For what its worth we have 2 citrix servers ( which are being used just for published desktop sessions ) and we are using group policy to redirect application data, desktop icons, and my documents to their network drive to anybody on those servers. For us that sped up logon and logoff times and general performance of the server / sessions
Posted via [H] Mobile Device
 
FileMaker pro has a PDF available with recommendations / instructions for FMP on terminal services. The following is for ver 8, ymmv.

http://www.filemaker.com/downloads/documentation/fm8_terminal_services.pdf

Also, I second folder redirection, AV exclusions (nothing for terminal server, but there are some suggestions for FMP), GigE switch for file transfers and general access.

User Guide that has antivirus information.

http://www.filemaker.com/downloads/pdf/fmp10_users_guide.pdf

There are Group Policy settings for restricting terminal services settings such as resolution, color, etc. These could also help you out.

http://technet.microsoft.com/en-us/library/cc780733(WS.10).aspx
 
So we've been able to force visual settings down to 8-bit color. Going to test with a few select users to see if they notice a difference. I'll post back soon with some results.
 
It seems that it is resulting in better performance. Looking into editing the GP to manually force a resolution.
 
8 bit color....ewww.

I restrict the server down to 15bit at the lowest. Anything less, I get complaints.

Also, disable audio.
 
8 bit color....ewww.

I restrict the server down to 15bit at the lowest. Anything less, I get complaints.

Also, disable audio.

How much of a performance impact will 15bit be over 8bit? We're currently at 15bit but I havent thoroughly tested both. 15bit seems ok. There are still hangups and delays, but it seems usable, however there aren't many people logged into our TS right now.
 
How much of a performance impact will 15bit be over 8bit? We're currently at 15bit but I havent thoroughly tested both. 15bit seems ok. There are still hangups and delays, but it seems usable, however there aren't many people logged into our TS right now.

No idea. With it set to 8 bit, the shop software is difficult to use, random shadows/odd colors/shades...
 
Are you running 32bit Enterprise Ed for >4GB memory support or 64bit OS? There are definitely quite a few known issues with TS/Citrix scaling running on a 32bit platform due to kernel memory starvation (among other things) that can result in Windows swapping constantly even when there is still free physical memory.

Also as someone asked earlier what kind of drives are you running and what kind of raid controller (IE: RAID5 or 1, Write Cache enabled/disabled)? Having run a number of fairly decent sized farms with TS and Citrix (50+ servers with 4000+ concurrent users) I can say that underpowered drive setups are a very common cause of performance issues.
 
We actually run a single 80gig drive without Raid. Would running 2 WD Caviar Blacks in RAID 0 be a worthwhile upgrade? We don't have the budget for something like raid SSD's considering we may need to upgrade to server 2008 64bit with CAL's. Would we see a significant performance increase?

Also, would disabling the page-file yield a significant performance increase? As per the post above, that would eliminate the memory swapping, would it not?

Thank you and I appreciate all of your input!

Edit: We are running 32bit Enterprise Ed for >4GB btw
 
Last edited:
Back
Top