A desktop virtualization company called “Wanova” came out of stealth mode yesterday. Their forthcoming software product called “Mirage” is a solution for streaming, managing, and bi-directionally syncing Windows disk images across a WAN. Their solution works at the disk driver level, so they don’t care if the client is bare metal / physical or whether there’s a hypervisor or VMM in the mix.
Conceptually this is a little bit like a combination Citrix Provisioning Server and application streaming. (Provisioning Server because it handles the entire disk image, and app streaming because it uses the disk locally and streams missing bits on-demand.) Of course there are some differences too, like the fact that Wanova streams in both directions (so user changes and data are streamed up to the server, meaning that everything’s always in sync.
Wanova’s approach to desktop virtualization?
It should come as no surprise that Wanova thinks server-hosted / thin-client VDI has a lot of limitations, like the fact you can’t use your desktop when you’re offline and there are performance issues with some apps via remote display protocols over the WAN. And they don’t like client-hosted desktops because you have all this data that’s out there that you need to manage. So this is where their product comes in.
Wanova talks about two main parts of their solution:
The is they built their own layerization capability for Windows disk images that allow them separate Windows into a base image, a machine settings layer, and a user settings layer. This concept is a critical for them as it allows them to update base images on-the-fly out in the field without blowing away everything the user customized or installed. And it means their synchronization engine doesn’t have to sync the entire disk, but just the changes the user makes.
In a lot of ways Wanova’s layers are the same as any of the other twenty or so desktop virtualization vendors’ layering capabilities, although of course Wanova’s layers were specifically designed and optimized for streaming. I guess my point is that I understand why Wanova rolled their own layerization solution, but I’m not impressed by it. I mean it’s fine, but whatever—a lot of people do that now.
The second part of Wanova’s solution is the cool part, and that’s their disk streaming and syncing stuff. They think the concept of “checking in” and “checking out” a disk image is kind of crazy, so in their case their disks are always in sync. If anyone’s used one of those cloud-based file sync services like Drop Box or Live Mesh, it’s kind of like that, except for your whole desktop (and that it’s Windows only). But you can imagine that combining this WAN-based sync with the layerization and you end up with a scenario that doesn’t actually have *that* much data to move around. (Their agent also handles compression, de-dupe, etc.)
I guess the other big difference between Wanova and the other companies out there is that Wanova’s product was specifically built for desktop virtualization. It’s not like they have a general purpose virtual disk format that they’re trying to shoe-horn into this system.
How it works
From a technical standpoint, there are really only two components that make Wanova happen: a filter driver and a Windows service. The filter driver tracks file system access and modifications and handles cache misses when a read is needed that isn’t local. The service does most of the heavy lifting for them, including the client-side network optimization, deduplication, streaming, the bi-directional updates, continuous data protection, communication, restores, etc. There’s also a handful of “little things” that tie everything together, like a tray app, sysreport utility, a boot driver, etc.
Since they’re just changing the way the disk is presented to Windows, they’re completely hardware agnostic (just like Provisioning Server).
Ideally you’d boot -time clients on a fast connection, although it’s actually possible to do that via the WAN too. The client prefetches a minimal working set while continuously streaming in the backend. (And of course it will fetch on-demand any missing pieces that are requested.) But once the data is down, it’s maintained in the cache on the local disk which means it doesn’t look that different from a normal installation.
It definitely seems that Wanova has the potential to solve a good problem in our industry. The only problem I see is that they have a specific product for a specific use case—it’s not really like you’re going to use Wanova for your desktop virtualization solution across the board, rather, you’ll use them to manage your offsite users only.
From a competition standpoint, I think a company like MokaFive jumps into my mind .
So we’ll see. I haven’t come across Wanova in the real world yet. I know they will be at VMwareworld in a few weeks, so hopefully I can find out more there.
What are your thoughts? Does this have potential? Has anyone seen Wanova in the wild?