Solved

Windows 2008 Terminal server RDP connections very slow

Posted on 2014-12-30
18
223 Views
Last Modified: 2015-06-07
There are about 20 computers on the network that all access a remote terminal server using RDP. This remote TS hosts all the applications and data that the business needs to function. Local computer use is negligible; perhaps some web surfing at lunchtimes, but nothing else except redirected print jobs from the TS going to the local printers.
The business also has a VoIP telephone system on a separate network, with its own internet connection. The phones have been no trouble apart from one glitch which was an issue at the remote server end.
There are two BT Infinity 80Mbit fibre connections on site, one for the TS connection and the other for the phones. Both connections run quite close to what it says on the tin - 74Mbit down and 18Mbit up, according to Ookla's speedtest and BT's own tool.

Here's the problem:

Performance on the hosted platfom is dire, alternating between bad and very bad. There is frequently noticeable lag when typing, but it can get as bad as 10-15 seconds between keypresses and characters appearing on screen. Anything involving moving images is basically unwatchable because it's so jerky, and  the audio is almost invariably out of sync. TS sessions often drop, though they usually reconnect fairly quickly. However, it has to be said that every so often things will run smoothly for a few minutes, but then it's business as usual again. Whenever things on the hosted platform are running badly, local internet access is unimpaired.

I've tried swapping the internet connections, and swapping routers, with no improvement. Likewise, enabling and disabling QoS had no effect. The routers being used are a Cyberoam CR25ia and a Draytek 2860, configured so that the computer and phone LANs and WANs are isolated from each other. Having everything on the same LAN with load balancing of the WAN connections was also tried, but that just made the TS sessions even more unreliable. The lines have been tested by BT and no faults have been found.
The hosted platform provider is adamant that it's a local problem, but I'm not convinced, particularly in view of the fact that local internet access always works well regardless of how badly the TS sessions are behaving.

I'd be very glad of input from those wiser than I...
0
Comment
Question by:Perarduaadastra
  • 11
  • 5
  • 2
18 Comments
 
LVL 37

Expert Comment

by:Neil Russell
ID: 40523362
First and foremost what is the specification of the terminal server?
CPU's?  RAM?  Physical or Virtual?
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523387
The terminal server is a VM. The other information I'll try and discover. However, I do know that the provider has increased RAM and CPUs more than once.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523426
Well, that was easier and quicker than I expected...

It's a fully 64-bit environment that currently has 27GB of RAM and three quad-core CPUs allocated to the VM. I'm also told that SSDs are used for the system volumes and that the data is held on a high-performance SAN.

Performance is still rubbish though!
0
Optimizing Cloud Backup for Low Bandwidth

With cloud storage prices going down a growing number of SMBs start to use it for backup storage. Unfortunately, business data volume rarely fits the average Internet speed. This article provides an overview of main Internet speed challenges and reveals backup best practices.

 
LVL 37

Expert Comment

by:Neil Russell
ID: 40523438
Can you do tests of performance when there is only 1 user on it, 5 users on it, 10 users on it tc?

Without knowing exactly what you are running on it its hard to say if the resources are sufficient but it has a reasonable start.

Bear in mind that IF by " three quad-core CPUs allocated to the VM" you mean that the VM is configured as having 3 Virtual Sockets with 4 cores per socket, giving 12 virtual cores, then I would say it is configured incorrectly.
 
I would be surprised if this is a network issue.
0
 
LVL 37

Expert Comment

by:Neil Russell
ID: 40523455
Other factors to consider are around the Host that the VM runs on.  How many CPU/Cores does it have in total?  How much physical memory in total? How many VM's run on it in total?  How many vCPU's are allocated in total? How much memory is allocated to VM's in total?

Is this hosted on VMWare or Hyper-V?
Do any other VM's on the host experience any performance degradation?

Can you give me a list of applications being run on the TS?
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523482
Unfortunately there are layers of administration and responsibility here, which is why I was surprised to get information as quickly as I did.

The hosted platform provider actually outsources the hardware and management to a third party that owns and operates the UK data centre where the platform is located. Getting more detailed information is going to be difficult, if not impossible.

The VMs are hosted on VMWare. The platform provider says that this particular installation is the only one that has problems. I have no information on the VM host hardware, or how many VMs it services, nor do I know how the resources are distributed, and I'm fairly confident that I won't be told, either.

The business itself is an accountancy practice. The running applications are MS Office 2010, Outlook and Exchange (I have no idea how Exchange is implemented, but I would expect it to be on a separate VM, and email sending and receiving works well), and a fairly heavyweight accountancy and payroll package that seems to do just about everything except make the tea. Four of the computers have document scanners attached that upload their scans to the hosted platform via FTP, but this seems to work well.

I'm interested in your suggestion that the configuration may be incorrect - if the one you describe is wrong, what should it look like? I will try and obtain more information, but I'm not optimistic.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523488
Tests have been run with varying numbers of users, but the results suggest that there isn't much difference between few and many.

I was on site yesterday evening when everyone had gone home and opened a single TS session, and its performance was very poor - unwatchable videos on Youtube, etc., so I'm inclined to agree that it isn't a network deficiency that is to blame.
0
 
LVL 37

Expert Comment

by:Neil Russell
ID: 40523510
I would suggest that you start by looking at task manager and perfmon stats on the server itself.  Only problem is that is it is the host that is throttling the guest VM you wont see that.

27GB might sound a lot but when you have 20 users on running all those apps id would be like a 512k XP workstation (Well almost).
12 vCPU's  configured like that can, depending on VMWare version, have a detrimental impact on performance.  Without access to the host performance info though, you are going to have great trouble proving anything.

If you have full admin access to the TS then I would suggest you start with perfmon and looking at what it gives you when you put it under a load. Video is a great test of remote connections and I would suggest you don't expect TV quality.
RDP is not meant for watching remote videos.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523540
I may be able to access the task manager, but only in read-only mode, so to speak. The server itself (physical and virtual) can only be tweaked by the data centre admins. I already know how difficult it is to prove anything in this situation - because of it the problem has dragged on for months, with no resolution in sight.

There are 20 computers, but at present there are only 13 in use.

I appreciate that RDP isn't designed for viewing remote videos, but I mentioned it because it was a way of easily showing just how bad the performance of the connection is. The real issue is the terrible performance of the link when it's being used in its intended manner, that is, sending keystrokes and updating the display.

What vCPU configuration would you recommend, to improve performance?
0
 
LVL 11

Expert Comment

by:rharland2009
ID: 40523548
How big is the pipe that connects your location to the datacenter? Keystroke/refresh lag can just as easily point to network latency.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523566
A BT Infinity 80Mbit fibre connection is used. Latency varies between 13-22ms, based on pings to the datacentre firewall.
0
 
LVL 37

Expert Comment

by:Neil Russell
ID: 40523575
"A BT Infinity 80Mbit fibre connection is used"

So this is over a public internet connection? Do you have a point to point VPN configured then between your firewall and the datacentre or TS ?
0
 
LVL 11

Expert Comment

by:rharland2009
ID: 40523596
Skimming over your previous comments, it sounds as if other types of traffic - even going to the same virtualization resources - are not affected by this terrible latency. Is that correct?
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40523683
@Neilsr:

Yes, it's a public internet connection. A VPN was initially used for the printer traffic, but this proved to be very slow and unreliable as well as giving poor print quality, and so was abandoned. In short, no, there isn't a VPN.

@rharland2009:

The actual delivery and receipt of email seems fine; typing them into the hosted Outlook can be a problem. Printing works well, and I've had no reports of delays with the FTP uploads of the scans, although this may be due the staff not paying any attention once they've hit the button. It's only traffic that uses the RDP connection to the TS that suffers. As I said earlier, local internet access always works well no matter how slow the TS connection is.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40524713
My comment about routers in the question is a little ambiguous, so I'll clarify:

The routers have each been configured to manage both the computer and phone networks, keeping the LANs separate from each other along with their respective WAN connections. There is no WAN load balancing or failover. Regardless of which router is actually deployed, the RDP connections remain as poor and erratic as ever.
Currently, the Cyberoam is managing the computer and RDP connections, and the Draytek is handling the VoIP phones; the LANs and their WAN connections remain entirely separate from each other. Unsurprisingly, the problems with the hosted platform connection are unchanged, lending further weight to the argument that something is amiss at the TS end of the wires.
0
 
LVL 15

Author Comment

by:Perarduaadastra
ID: 40544215
Update...

Because the hosting company is trying to blame the routers, I've tried the standard BT Business Hub 3.0 which gave terrible results, and an elderly Netgear FVX538v2 which didn't seem to be any different from the Cyberoam and the Draytek units, except that audio over RDP worked a little better!

However, because of the long history of poor RDP performance the hosting company is now talking about using a leased line to make the problem go away, which may or may not be a tacit admission that they don't really know why the connection has been so bad for so long...

I'll post back with any new information.
0
 
LVL 15

Accepted Solution

by:
Perarduaadastra earned 0 total points
ID: 40808374
Well, the leased line was finally installed, and bingo! All the problems went away. The number of hops between the company's router and the data centre firewall has dropped from 18 to 5, and ping times have fallen to 2-3ms.

I can't say for sure what the problem was, but I wouldn't be surprised if it was a fault or misconfiguration on one of the 13 hops that is no longer used by the connection. In any case, all's well that ends well...
0
 
LVL 15

Author Closing Comment

by:Perarduaadastra
ID: 40816717
Those who contributed did their best with the very limited information I was able to supply about the hosted platform, but the lack of information I was able to obtain meant I couldn't give them anything to work with.

In the end, I could only confirm that a leased line fixed the problems, and even now the reasons for those problems remain unclear.
0

Featured Post

Simple, centralized multimedia control

Watch and learn to see how ATEN provided an easy and effective way for three jointly-owned pubs to control the 60 televisions located across their three venues utilizing the ATEN Control System, Modular Matrix Switch and HDBaseT extenders.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Suggested Solutions

Title # Comments Views Activity
Solar Winds can't see SQL Server Express 17 52
IP range 6 58
AD Design Best Practices 6 32
WDS can't PXE boot 3 26
In the hope of saving someone else's sanity... About a year ago we bought a Cisco 1921 router with two ADSL/VDSL EHWIC cards to load balance local network traffic over the two broadband lines we have, but we couldn't get the routing to work consi…
Most of the applications these days are on Cloud. Cloud is ubiquitous with many service providers in the market. Since it has many benefits such as cost reduction, software updates, remote access, disaster recovery and much more.
Here's a very brief overview of the methods PRTG Network Monitor (https://www.paessler.com/prtg) offers for monitoring bandwidth, to help you decide which methods you´d like to investigate in more detail.  The methods are covered in more detail in o…
In this tutorial you'll learn about bandwidth monitoring with flows and packet sniffing with our network monitoring solution PRTG Network Monitor (https://www.paessler.com/prtg). If you're interested in additional methods for monitoring bandwidt…

680 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question