Steelhead Appliance

[ugur]

RDP alone cannot cache the files. Each time one requires a file, it is downloaded from center into edge. This means inefficient bandwidth utilization. However, when RB is configured to optimize RDP traffic, it downloads files once into edge reducing redundancy with built-in mechanisms so that end users do really benefit from this.

A 438 KB files takes 55 seconds to download with native-RDP on a 256K connection for all the times. However, when RB optimizes RDP, it takes 40 seconds on the first run, 0 seconds on the consecutive runs.

This is a real benefit over native-RDP.

Regards.

Ugur Unluakin

[joshtseng]

A former employee of a vendor that claims to "accelerate" Citrix ICA traffic once told me that to generate the performance improvements for the thin client application, what they do is run a recorded script featuring a series of ICA operations through their WAN optimizer boxes. They then run the same script again, a second time, and measure the compression benefit. Because it's a recorded script featuring the same operations, with ICA sending the same bytes on the wire, their WAN optimization devices are able to store the bytes and record some amount of data reduction. The test results are good enough to create a marketing brochure for, but this is hardly a practical test result that can be applied in real life...

Josh

[amh]

Working for a Riverbed partner i got curious myself, especially since a competitive product claims to be able to optimize ICA, at least in certain circumstances. If we disabled encryption and compression, and use drive mapping or printing over RDP or ICA, we see optimization on the print jobs and file transfers, when tunneled over TCP port 1494/3389, but there is no effect whatsoever on the screen refresh rate or byte count for the "normal" thin traffic, which makes the previous replies perfectly consistent with our findings.

Cheers,

Arne

[bgilbert]

Check out my post here http://www.wdsforum.org/forum/index.php?sh...&#entry1001

[bgilbert]

Hi Vijay,

RDP traffic is passed-thru the Steelhead by default. You can change this so RDP is intercepted by the Steelhead, but RDP inherently comes with pretty good compression and there is really no layer-7 latency optimizations that can be performed on this traffic.

The best way to handle this traffic is to leave is as pass-thru and let the Steelheads do what they do best; make more room on the pipe for this traffic. You can also use the Steelhead QoS enforcement facility to give the RDP traffic a priority as it enters the queue.

Bob

[vijay]

Hi Bob/Blanco,

Could you please throw some light on how RDP is handled?
We do have huge amount of RDP traffic over our WAN.
Can SH appliances perform any optimisation for RDP traffic in terms of data reduction or atleast BW usage reduction by compressing the RDP traffic?


-VJ

[duke131]

Blanco and Bob,

Thanks very much for the great article and insight. It all makes sense.

Wen

[bgilbert]

There are three cases where Steelhead appliances can potentially help ICA traffic:

1. By making more room on the pipe in congested environments where Citrix sessions are fighting for bandwidth, response time and quality of connection can improve.

2. Citrix print traffic can be bulky and slow over WANs and Steelhead optimizes this traffic, reducing the bulkiness and ultimately reducing how long it takes for these print jobs to complete

3. In a congested environment, the Steelhead's QoS capability can allow Citrix traffic to be given a higher priority over the other traffic going through the same pipe.

The net-net is that if you already have Steelheads in place and your general Citrix operations remain slow, you would probably get no additional benefit from adding ICA as traffic that the Steelheads will try to optimize. The Steelheads should be taking care of congestion so that would probably not be your bottleneck. Again, if print is your pain, then I would suggest turning on ICA interception.

Bob

[blancolam]

Wen,

The reason why the Steelhead pass through the ICA/RDP traffic is because ICA in itself is a very lean protocol already. Click on the following link to see an article published by Citrix themselves explaining the ICA protocol.

http://support.citrix.com/article/CTX113505

Regards,

Blanco

[duke131]

I understand that Citrix and SSH (RDP) are passed through the Steelhead at the moment. Will these eventually be optimized?

Thank you.