While there's not too much that most readers of BrianMadden.com agree on, I'd be willing to bet that we can all agree on (1) desktop virtualization will be big, and (2) no single vendor today offers anywhere near the complete nirvana solution.
I've written quite a bit about how I think that all of the technical components we'll need for our nirvana solution will be available next summer (and I still think we're on track for that). The problem that's rapidly beoming clear is that the entire "solution" (at least as it stands today) will require each customer to cobble together products from about 15 different vendors. This is so complex and so risky that it means customers won't actually do that, and will instead just leverage the minimum number of vendors as possible to solve very specific tactical problems.
So that's fine, but it means that we're not going to get to the nirvana solution because it's too hard.
My point is that of the 100+ vendors in the desktop virtualization space, probably three-quarters of them are "one hit wonders," where they do one specific thing well (really well!), and the rest of their stuff is not interesting or not as good as another vendor or something they just threw in last minute so they could call their product a "solution."
Let me give you some examples. This list is in random order and only covers about 5% of the one hit wonders out there:
- Wanova, a new startup I wrote about last week, offers streaming disk images across a WAN. I'm sure they're very good at that and maybe better than anyone else in the industry. They also offer layering of Windows which I'm more nervous acount.
- Atlantis Computing has that really cool disk I/O virtualization, which completely rocks and I think should be standard in every VDI deployment. They also support application-specific layers, which I fear are very brittle. So with Atlantis, I want to use their dynamic disk composition thing, but I'll leave the app isolation and virtualization to another vendor.
- Neocleus seems to have the best client hypervisor on the market today (since they can sort of combine/bridge the emulation, paravirtualization, and pass-through techniques), but they also threw in a basic management engine which I'm not impressed with at all. I'd love to see their hypervisor everywhere, but managed by something else.
- MokaFive has done an amazing job digging into the innards of various VMMs and getting them to boot from USB keys, enabling cross-platform suspend/resume (suspend on VMware Workstation for Windows, resume on Fusion for Mac). They also offer Windows layering. Are they the best at that too?
- Kaviza has a really cool self-configuring virtual appliance that essentially lets you build a desktop grid. So if you need more capacity you just add their virtual appliance to a new server, point it to your existing environment, and walk away. Their system automatically figures out what resources you need on the new hardware and starts copying stuff over and using it (connection broker, disk images, VM hosts, etc). It's very cool. But do I want to use Kaviza for my connection broker too? Do I want to use their disk image management?
And that's just a partial list among the startup-type vendors. Think of the client device. Think of the display protocol. Think of the user environment management. Think of the printing. In every case, there's probably one vendor that pops in your mind as the "goto" vendor for each of these.
Would it be technically possible to combine all of these, pulling out just the best parts to create a nirvana solution? Maybe. But it wouldn't be practical. First of all it would be waaaaaay too expensive... probably pushing $2,000 in software licenses per user. Second, it would be a nightmare to support. And third, a lot of these products are built so that you have to choose them for everything they do. I don't think I have the option of using Wanova for WAN-based disk streaming but replacing their layering with MokaFive's. I don't think I can use Kaviza to automatically manage my entire environment while replacing their connection broker with Leostream's.
So the reality is that I have to make choices. I have to decide which issues are most important for me, and then select the vendor that best solves my specific challenge. And if that vender requires that I use some other component of theirs that's not as good as something from another vendor, well, in most cases that's just too bad.
I wonder a bit about the long-term viability for these one hit wonders. Is it realistic that most of them will be acquired by the big vendors and integrated into their suites? Can they stay independent and continue to sell their single hits? Should they focus on integrating with others instead of being complete solutions? Of course this all depends on which vendor and which technology we're talking about, but overall they're in a really hard place, especially considering that a lot of people want to virtualize their desktops to save money.