Yesterday, Sumit Dhawan, the head honco of the Citrix Receiver and End Users Services Group presented at the Gartner Catalyst Conference along with Vittorio Viarengo, Vice President of End-User Computing for VMware in a session moderated by Gartner’s Chris Wolf entitled “Protocol Wars: What Matters”. Before I get to how PCoIP cured cancer and won the world cup, I wanted to reiterate the premise of Sumit’s presentation:
We at Citrix believe that the “winner” of the protocol wars is the one to deliver the best user experience at the lowest cost. But really, it’s not about the protocol to us. We stopped talking about ICA (our protocol) a long time ago because we realized that what people actually cared about was a solution that maximized use of the resources of the end-point, be it thin client, phone, tablet, laptop, or PC while reducing costs on the back-end. Does a protocol do that? Not alone. You need a tightly integrated solution that is invisible to the end user, aware of its surroundings, and makes use of every networking and computing resource you give it. That’s HDX – a technology purpose built to provide the Highest Definition User Experience regardless of back-end infrastructure, network conditions, or end-point device. You can call it a “protocol” if that makes you feel better, but it’s much more than that.
Which brings us back to Vittorio. His claim? Teradici has “removed of the final technical hurdle” that has been responsible for the lack of View adoption in the Enterprise. That hurdle he speaks of? WAN performance.
Let’s just take a pause here and reflect. What Vittorio is saying is that the protocol, the thing that they outsource to Teradici, has been the anchor that has stopped VMware View from beating Citrix XenDesktop in the Enterprise. And now that problem is mitigated. Just like that all the problems with PCoIP are a thing of the past, and now there’s nothing between View and VDI domination!
There two ways to address this claim: A business case response, and a technical response. I’ll give you both:
XenDesktop doesn’t beat View because of the protocol alone. How about our breadth of supported devices? Did you know View doesn’t even have PCoIP support for their Mac client? We have support for nearly every device imaginable because we are in-tune with the fact that IT isn’t always able to dictate what devices people bring to work, but they are expected to enable all employees to be productive. Receiver is the only client that enables IT to keep up with the growing spectrum of worker end-point devices. HDX is the only solution with unique optimizations for maximum end-point device resource utilization for the best end user experience, while XenDesktop provides administrators with single image management and the flexibility to provide the right desktop for the right user. Another aspect of the solution worth mentioning is that XenDesktop Platinum Edition ships with our WAN optimizing Branch Repeater product.
It’s because we’ve built our own technology we can dig deep into it and build mechanisms that not only accelerate its flow through networks, but also optimize all the other applications that utilize the same link. Our ability provide a complete holistic solution is why our customers tell us they choose us. The best VMware can do is send Teradici a wish list. Let’s just put it out there: VMware can afford to buy Teradici. Why don’t they? Because View is not that important to them. They can’t even give the stuff away. When I analyze our Win/Loss reports at the end of each quarter I see us losing opportunities not to View, but rather to abandoned View POC’s that we weren’t involved with – opportunities where VMware POC’d View and left such a bad impression that the prospect was soured on VDI all together.
I hate to break it to View fans but there’s still no good way of optimizing PCoIP. I’ve looked at the “technical” chart posted on Warren Ponder’s (Director of Product Management for VMware) blog about a hundred times and here is what I get:
1. Client side caching. Ok, how? “With new advanced capabilities, View will cache images and portions of the desktop composition to minimize retransmission of pixels across the network. This results in a significant reduction in the bandwidth of individual sessions – perfect for static screen content that does not need to be updated or in situations where only a portion needs to be refreshed.”
A good step forward, so kudos. However, unlike HDX there’s no calculation of data across desktops, meaning if you and I are watching the same video on a View desktop with PCoIP then our bandwidth utilization grows exponentially because there’s no optimization across desktops. Whammy #1 – Think about how this becomes a problem for 501 desktops across the WAN. Why did I choose 501? Bear with me, we’ll get there.
2. Lossless CODEC – “One of the leading causes of high bandwidth utilization for remoting protocols has historically been from advancements of fonts such as ClearType fonts. This has been well known for some time. One of the new enhancements for VMware View with PCoIP is an update to the PCoIP lossless CODEC used for text that will improve compression – resulting in lower bandwidth requirements.”
Again, a worthy accomplishment – congratulations on getting here. Now how about going beyond fonts like HDX and optimizing file transfers, audio, and apps? Someday perhaps..
3. The ability to build to lossless – “Many of you familiar with PCoIP know one of the biggest benefits of PCoIP is the rich user experience it provides and its unique ability to offer a fully lossless experience. For users who want to maximize bandwidth savings vs. lossless image quality, we will provide an option for disabling PCoIP build to lossless in favor of build to perceptual lossless. Providing this capability allows customers the option to choose the right solution for their individual deployments.”
If there’s one thing users hate it’s options. Why can’t the technology figure it out for me like HDX does? If you take a look at HDX monitor you can see all the variables HDX keeps track of while delivering desktops and applications to end-users. When a variable like network latency changes, HDX automatically reacts and adjusts not only the image quality, but also the very nature of the connection to ensure the best user experience. PCoIP is great at delivering one desktop to one user, the second more users pile on user experience for everyone starts to deteriorate.
Finally, let me touch upon this “Up to 75% reduction of bandwidth” thing.
I’ll start by telling you a little story we used to hear in POC’s when facing off against VMware View. They’d say, “Yes, PCoIP’s ability to consume large amounts of bandwidth is a good thing because you want to maximize the use of your pipe for delivering desktops.” They would attack HDX because it “artificially limited the bandwidth it utilized and didn’t take advantage of the bandwidth potential of the link.” Uh-huh. Let that simmer.
Back to 75%. Even if that number was correct (and by the way “up to 75%” means “usually much less than 75%”), the very nature of PCoIP, the fact that it is a UDP-based protocol means that it is its own worst enemy. Each instance of View over PCoIP adds to the aggregate sum of bandwidth needed to maintain the the optimal user experience such that 100 instances of View would need to use 100 times the bandwidth of a single instance of View in order to maintain the same end-user experience. In plain English: As View crams in a bunch of sessions on a bandwidth constrained network, user experience goes to Shinola. Add in real world network bandwidth limitations, a mix of protocol congestion, dynamic latency, don’t forget a pinch of packet loss and even with the new “up tp 75%” bandwidth reductions you’ve managed to clog up your WAN and piss off your end-users. And here is where I get back to 501. 500 concurrent VDI sessions is where we at Citrix draw the line between SMB and Enterprise desktop virtualization deployments. Can View support 501 concurrent desktops over your WAN? I believe the answer is no, and this is why I would like to conclude with the following:
VMware View does not compete with XenDesktop. It competes with Kaviza (and not very well there either). It’s an SMB VDI solution that is meant to work well for up to 500 users over a LAN. That is its sweet spot, and we look forward to competing with VMware in the SMB market with Kaviza. But XenDesktop is an enterprise grade product, built to enable strategic desktop transformation for companies who want to remain competitive in a world where employees can work from anywhere, on any device and demand IT as a Service.