I have been giving a lot of thought to client side virtualization lately, specifically the idea of type 1 hypervisors on x86 machines. After all this time I have decided that they aren’t worth any more brain cycles (I have few to spare). Back in 2006 I wrote about the client side hypervisor at Brian Madden’s site and was really excited about what I might see by 2009… What did all that excitement get me? Nada. Zip. Nothing. Client side hypervisors today suck. Their mgmt. tools aren’t any good, and they really don’t make IT’s job any easier.
A hypervisor solution for PCs that gains any traction will come from MS and will essentially be Hyper-V for desktops. Anything else is a waste of time for the foreseeable future. Could my mind be changed? Maybe. Could I be completely wrong? Sure. Obviously I was wrong when I wrote about my excitement for these clients in 2006!
Ron's basic timeline for the type 1’s:
- Talk started in the community around 2006
- By 2008/9 pretty much EVERYONE was SURE we would have type 1 client side hypervisors readily available today. Hell we thought it would be an option when you bought that new laptop!
- Citrix and VMware were going to be battling in that space and small startups were hoping they were far enough ahead of the big boys to get bought or get traction
- By 2010 VMware had given up on the Type 1 on the client and Citrix was dealing with what was basically a 3rd gen prototype that supported a handful of laptops.
- Citrix announces XenClient 2.0… everyone yawns
So, I am done with the Type1 on the client. I am sick of waiting for one that supports more than a handful of hardware configurations and doesn't choke on standard peripherals. Yes, client hypervisors are extremely complicated. Support for drivers and devices is a NIGHTMARE. I understand. And, yes I will probably get a bunch of emails telling me who has one now, and how it’s better than Citrix or supports more than I think... whatever. Don’t waste your keystrokes. Let’s face it: Even with the 2 or 3 that are available now, people aren’t really seeing a use for them. I see a few uses, but most organizations don't even see that… and I think I know why.
Companies get the idea of the hypervisor in the datacenter. Yes you have to patch and manage a hypervisor on that Dell or HP server, but you get to stack 20 or 40 or 80 VMs on it for your trouble. The price you pay is pretty small compared to the savings. But, on the client side, the story is completely different. You will basically have to manage a Type 1 for 1 or 2 VMs on that desktop. Then setup a management infrastructure to manage the hypervisor AND the Windows VM sitting on top of it. This management layer for Windows may (OR MAY NOT) be the same tools you use today. If it is the same tools, what the hell did you buy the hypervisor for!? If it’s not, well that is yet ANOTHER tool you have to add to your environment to manage a subset of the Windows machines. Oh, and as a bonus, you can’t get rid of the “other” management tools because the hypervisor you bought only supports like 8 specific models of machines.... talk about a pain. Why deal with it?
The other problem I am starting to see with the client hypervisor is the direction client devices are moving today. Much like all desktop virtualization, some in IT thought that we could just place a hypervisor on the client and benefits in administration and cost would magically rain down upon us from the IT gods…. I guess that is all that happened on the server side right? We dropped a hypervisor on our servers and magically your datacenter looked like that stupid IBM commercial right?
Wrong. There was a lot of thought about how virtual machines and hypervisors would benefit the datacenter and IT operations. The results of a consolidation effort were clear LONG BEFORE the technology was even ready for prime time enterprise usage. On the client side it seems like (at least to me) that those of us championing the client side hypervisor are force feeding a product into an environment that hasn’t even figured out what the benefits will be. It’s the exact opposite of the datacenter virtualization play.
And then, as you look at clients and how people work today, you see the iPads, Smart phones, book readers, etc etc. The market is dictating small, powerful, always changing, flexible devices. And most of these devices are not your typical x86 laptop that we envisioned the type 1 hypervisor sitting on. Hell, my wife is totally enamored with that commercial showing a phone w/ a keyboard & monitor. How many of you wouldn’t KILL to have a phone that could also replace your laptop? Think it won’t happen? Eventually? Sure it will. And where does the x86 client side hypervisor fit in that world?
Answer: It doesn’t.
These next gen devices may connect to a desktop somewhere (hosted at your home or your datacenter or wherever) and will have the ability to run a bunch of apps locally just like the iPads and phones of today. The user won’t need a local hypervisor, so why pay for the overhead (in hardware perf, management and license costs)? Windows or Windows apps will just be another app accessed by these devices… Windows (as far as these devices are concerned) basically becomes a streamed version of Angry Birds.
The Type1 Hypervisor is / was a pipe dream. It was thought that it could change Windows management and that was the real driver behind a lot of enthusiasm in our world. Managing distributed Windows desktops just plain sucks. All of the tools out there have every option you could imagine, and that makes them complex, and complexity leads to failures. But the Client Hypervisor isn’t the fix for that, therefore it has limited use to IT for the foreseeable future, and I am done with it.