(written by Mark Jimenez, Riverbed Technical Marketing Engineer)
“It’s like the High-Speed TCP protocol for RemoteFX.”
Today, we're pleased to be able to present Riverbed Steelhead optimization for Microsoft RemoteFX. This is a big day for both Microsoft and Riverbed here a TechEd EMEA in Berlin, in that Microsoft is publicly showing off their new RemoteFX technology, with Riverbed simultaneously demonstrating optimization of RemoteFX sessions.
Be sure to also read Microsoft's blog entry.
RDP 7.1 with RemoteFX enables a rich, local-like user experience for a virtual desktop infrastructure (VDI) by virtualizing the GPU on a Hyper-V server for Windows 7 virtual machines. With the bulk of the video processing now performed on the server, the remote user can experience full-fidelity video with 100% coverage for all media types and highly-synchronized audio, rich media support including Silverlight, 3D graphics, and of course Windows Aero. At Riverbed, we have worked with Microsoft to deliver a solution that increases performance of RemoteFX over the WAN.
At the Riverbed booth at TechEd, we are showing 2 laptops, both running Windows 7 and the SP1 RC Remote Desktop client (mstsc.exe). On the server side, we have an HP server running Hyper-V for Windows Server 2008 R2 (with SP1 RC). Both laptops connect to the HP RemoteFX server on 2 different Windows 7 SP1 RC VMs, through a simulated WAN (50ms, 10Mbps).
The optimized session resulted in a significantly more acceptable user experience.
Having the 2 laptops side-by-side clearly showed the difference between unoptimized/unusable and optimized/usable remote user experience, not just for watching video, but also interacting with common productivity applications (Microsoft Office, Internet Explorer, etc.).
Prior to TechEd, we tested a single user scenario under various latencies. In summary, latency drives down maximum throughput, which results in fewer Frames Per Second (FPS). This still happens with Riverbed Steelheads, but at a more acceptable experience for the user (see graphs below). These numbers are expected to change depending on a variety of conditions, including bandwidth, round-trip time delay, packet loss, and user workflow.
Further testing will be the basis of a future deployment guide to more comprehensively describe how these variables will affect user experience.
Disclaimer: Any performance data contained herein refers to tests conducted jointly with Microsoft with the current RC version of product (as of November 8, 2010). A more comprehensive whitepaper will follow to demonstrate how this solution scales with various network conditions.