Around 20 years ago, Vanilla Ice – “Ice Ice Baby” was the hippest song on MTV, and 14.4 K modems were still the fastest way to connect to remote computers (the internet was still in infancy at that time). Citrix engineers were busy working on optimizing their remote connect protocol (ICA) because they knew, even at that time, a great user experience together with great performance is the key to making a delightful product.

Unlike our friends at VMware (Build-to-lossless feature), we at Citrix don’t think you have to compromise performance in order to deliver a great user experience over any network. We have continued to push the boundaries and innovate. With HDX technology, our customers all over the world get the best user experience with their virtual desktops no matter where they use them – in the office, from home, or in branch offices.  We think that is why people love their XenDesktop virtual desktops, because they work anywhere, just like people do.

A couple of months back, VMware published a report not approved by Citrix (more on this perhaps in my next blog) on performance tests they commissioned that evaluated a single user using single virtual desktop on a “quiet” LAN to support their claims that PCOIP provided equal or better performance than HDX.  What experienced IT professional is going to test a single user environment and extrapolate that to expected performance with hundreds or thousands of people in departments or divisions?  Certainly not any IT pros we work with.  Maybe that single user test is important for the typical View deployments?  But seriously, the report does not simulate any production environment and has little value in real world deployments.  Also, in the same report the “Best” performance is in the multi-media redirection scenario (which our customers use in real world) – and is won by XenDesktop but VMware simply tries to discount this very relevant use case.

After we stopped shaking our heads in disbelief, we saw this as an opportunity to provide valuable virtual desktop user experience guidelines to IT professionals.  We were working with to the same test vendor (Principled Technology) who did the testing for the VMware report. We asked them to test XenDesktop and View again, but this time with many users (up to 100) over realistic WAN conditions, with the goal of providing meaningful, real-world virtual desktop design guidance.  We know WANs, but we wanted to get additional third party guidance on scalability and performance over WAN, so we engaged ESG to better understand two things:

  1. What does real-world WAN traffic look like for remote and branch offices?
  2. What else can we learn to improve performance over WANs?

The ESG research found out that typical WAN traffic consists of email, office application and SharePoint data with an average 75ms latency and 1-2% packet loss. This provided us with balanced guidance on the WAN conditions for Principled Technology to use in the re-test.

Principled Technology tested both products out of the box for baseline comparison, XenDesktop (no surprises here) out performed View in every category. However, we also wanted to make sure we gave View a fair comparison, so we tuned View, based on VMware’s best practices white paper, and also compared it with the baseline XenDesktop environment. The results, XenDesktop with default configuration still performed much better than View tuned following most of their best practices.

Now, we also know that WAN optimization can also makes a huge difference in virtual desktop performance; that’s why we offer Branch Repeater VPX as a feature for XenDesktop.  Branch Repeater allows you to control traffic flow, accelerates HDX traffic and improves the quality of remote virtual desktop sessions. So we also tested XenDesktop with Branch Repeater compared to View optimized for bandwidth.  We observed that XenDesktop delivers twice as many virtual desktops, on the same WAN link with superior user experience than VMware View. Customers can reduce WAN costs and don’t have to look for third party WAN optimization tools, not that it would matter in case of VMware View.  Third party WAN optimization products cannot accelerate PCOIP, being an encrypted protocol.

In order to publish this report, the VMware EULA requires us to send them a copy of the report for approval.  We have.  We also recorded these 2 minutes videos doing side-by-side comparisons so you can clearly see how the two products performed.  Look for the link to the videos at the end of the report.

View 5.1 wasn’t available at the time Principled Technology conducting these tests. However, according to VMware release notes there hasn’t been any significant improvements done for PCOIP.

The net is that when you are conducting POCs for  testing VDI or Desktop Virtualization solutions. You need to test things at scale and test solutions under real world conditions vs testing single user or single desktop under perfect condition.

Download the detailed report.

Reach me on  Twitter for any questions – @vishalg