Peter Christy

Citrix Drops a Fascinating "VDI" Shoe

In the context of disclosing a partnership with Intel, Citrix pulled the veil back on their plans for desktop virtualization. To date the "VDI" solutions (including Citrix' XenDesktop) offered great potential for PC support cost reduction and security but a high price in hardware cost, performance and even basic functionality. Running PC applications on a server increases the cost and decreases the performance. None of this should be surprising because the basic value proposition of a PC is the tight coupling between the processor and the display, and that's exactly what these solutions give up. In our December Desktop Delivery report we noted all that and pointed out the same benefits could accrue if the user's applications were run in a virtual machine on their PC (not on a server) at lower cost (really using that PC for something other than a thin client), at much higher performance (everything local) and with greater functionality (since it still works if you're not on the network). Much to our pleasure that's exactly what Citrix concluded as well.

Citrix brings strong assets to the effort because of their acquisition of XenSource and close relationship with the OpenSource Xen hypervisor project. Citrix reasonably argues that a PC hypervisor and a server hypervisor aren't the same beast. The Xen hypervisor that Citrix will use here is designed to give direct control of some PC hardware to a virtual machine. In a virtualized data center the primary I/O is to the network and to storage, both of which must be kept secure and controlled, so giving direct access to hardware has limited practical benefit. On a PC the issues are very different. Giving direct access to graphics hardware (especially given interfaces such as Microsoft's DirectX) has very high benefit in terms of graphical and interactive performance, and essentially no risk. Citrix also argues that Xen was designed with much more attention to the secure isolation of the different VM's, something that wasn't as important on the server. In a virtualized data center there are still strong controls on what software will be run because it runs inside the network and with direct access to sensitive data. On a PC, again the situation is very different, since a clear goal is to be able to run riskier software sometimes, and a separate virtual machine is a great way to do that as long as the hypervisor protection is strong (alternatively it's a great way to run sensitive and secure software like business applications with protection). So Citrix' claim that there are a lot of different choices in a PC hypervisor has good support.

To do desktop virtualization directly on the PC requires all the mechanism needed to do it on a server (for example the image synthesis and management tools already in XenDesktop and the competitors) and then some. On the PC you also need to keep all the virtual machine images encrypted and secure, you need a way to update them as soon as possible and you need to put a little more mechanism down on the PC. But the investments and tools needed to do PC virtualization on a server all translate nicely when you run a managed VM directly on the PC.

So where, you might ask, does Intel fit in? The partnership with Intel is how Citrix is building bridges to the PC OEM's so that enterprise customers will be able to buy PC's with the hypervisor preinstalled under the normal software load. Citrix is happy to make that hypervisor broadly and cheaply used for all purposes including ones that might be viewed as competitive with Citrix' own efforts —  the more PC's with bare metal hypervisors the better. It's also worth noting that Intel has been a strong supporter of hypervisor developments and a key contributor to the Xen project from the beginning.

And how does Microsoft fit in you might ask? Although MS has hypervisor technology they have no public plan to ship a client hypervisor. Since the assumption must be that the hypervisor is essentially free to the PC user, having a client hypervisor would be important to MS if not having one diminished their ability to sell Windows licenses, or if a competitive hypervisor threatened their ability to control the O/S stack. But neither seems to apply to Citrix, a long time business partner and a technical partner on MS' own Hyper-V efforts. Microsoft too has been a Xen supporter because it weakens the proprietary control of VMware.

The interesting open question is where and how VMware will weigh in. They have discussed similar plans in a similar late 2009 timeframe. They haven't yet announced a partnership with Intel but that wouldn't come as a surprise. The open questions are what kind of hypervisor they deliver and how open and inexpensive its use is.

Cisco Video Architecture

At the C-Scape Analyst Meeting, video was the hot item for the obvious reason that increased video use places great demands on network infrastructure and drives network investment. As part of this, Cisco talked about a broad network "architecture" called medianet as the underpinnings. I had a chance to follow up with a delightful hour with Guido Jouret, the CTO of the Emerging Technology Group. You have to be wary about claims of network "architecture" because in the past many have turned out to be more "marketecture," but in the case of video there definitely is meat. When today's Internet first appeared as the ARPAnet the "backbone" links were 56Kbits/sec and the target application was command line use of remote computer systems. It was inconceivable then that any form of audio other than telephony would be served, much less something like HD video (survivable telephony was the initial driver for packet network development). Over the years, technology has obviously made huge advances and we find great value in network convergence —  putting more and more network traffic onto a fabric made with common technology (TCP/IP). To make video work, it isn't enough to just pile on bandwidth, or more accurately, we're many years or even decades from the point where excess provisioning might be a solution. Unlike most network applications, there is no graceful fallback for video when network performance diminishes. Quickly the video experience becomes totally unsatisfying. Cisco has a broad set of technologies meant to address video issues, some already in place in their IP/TV offering and broader product lines. The critical new additions are (a) video devices and subsystems that self-identify their capability and needs; (b) real-time management of network performance and the ability to intelligently respond when problems happen (today if there are four concurrent TelePresence sessions sharing a common link and the available bandwidth goes down under the collective requirement they all hang up, even though any three could continue without impact, or all four could continue if they could agree on acceptable fallback conditions); and (c) the ability to place critical video processing (e.g., video transcoding) at the most intelligent place in the network (e.g., it doesn't make a lot of sense to carry a high-bandwidth signal all the way across the network and then down convert it at the receiving end if it could be down converted much earlier and all that bandwidth saved). Video over TCP/IP is at best very difficult and very expensive compared to most uses of these networks, so exactly what will go over the Internet and what will stay on cable or satellite is still very much an open question in my mind. Cisco is putting in place the technologies needed to make it work as well as possible and enable the applications that do make sense. Along the way they are finally developing a real idea of what the "network as a platform" really means (most of the earlier versions were unquestionably primarily marketecture). It seems clear that 2009 will start to paint a much clearer and exciting picture of how video can exploit TCP/IP networks.

Healthcare as SaaS

Obama speaks intelligently of making economy accelerating investment with long-term benefit as a goal, and has highlighted Healthcare IT specifically. We applaud this perspective and believe that this is the right time to strongly push medical IT as a service. SalesForce has invested both profit and time supporting good causes, and in doing so helped create a set of municipal and educational applications that demonstrate clearly why most municipalities and school districts should get out of the IT business as soon as possible because most of the apps they need can be provided as a common service, better and at much lower cost (IT is not and should not be a core competence). The same is true of healthcare. We spend more per capita on healthcare than any other country and get at best incomplete and often mediocre healthcare as return. A lack of comprehensive data repositories is a critical problem, both for the cost-effective delivery of high-quality care, and as importantly for making intelligent decisions about what healthcare should be available to everyone (what makes sense). The HIPAA regulations have driven progress but at a glacial, governmental pace. Isn't this the time to reflect the SalesForce experience, and to use our cloud computing experience to create shared healthcare record repositories and SaaS services? A national healthcare record system is either a massive and daunting IT problem, or an insignificant use of either Google's or Microsoft cloud resources. I'm not one to a tout IT technology for the sake of IT technology but it surely feels like this is the right time for major change.

Fry's Loses Big

We've spent a lot of attention in the last few years to issues of "business governance" and "regulatory compliance." These often seem like overbearing oversight but then from time to time the significance of the issues jumps out. Take Fry's for example, an interesting and unique retailer that employs lots of low wage workers selling high tech. Just like someone running a coffee shop or bar, there are lots of processes in place to make sure these employees and their customers don't skim a little off the top to supplement their meager wages. I recently returned a hard drive in the factory shrink wrap, and discovered that Fry's removes the shrink wrap to make sure that the devious customer hasn't neatly concealed a scam of some sort. There are lots of checks and balances to make sure that nothing easy gets through. Unless of course you're the VP of merchandising and have a gambling problem in which case it seems roughly $100M slips through the cracks. Fry's isn't a public company, and if it was it's hard to imagine that this could have happened because clearly their "business governance" had holes large enough to drive large trucks full of cash through. You have to wonder whose brother in law was auditing the firm and whether they even had a CFO. Regulatory compliance is just a big word until you see something like this up and close.