Yesterday VMware announced that they acquired CloudVolumes for an undisclosed amount. Originally called SnapVolumes, CloudVolumes was founded in 2011 by Matt Conover, Shaun Coleman, and Matthieu Suiche, (and friend of the site Harry Labana joined them this past April after writing several blog posts on BrianMadden.com about them over the past two years).
CloudVolumes is a software company with a product in the Windows application delivery / containerization / "layering" space. Essentially they let you take a before and after snapshot of an application when you install it, and all the changes are written into a VMDK or VHD disk image that can be mounted (even when a user is logged in) to a machine. Once mounted, the CloudVolumes agent hooks the disk image into the disk chain and all the files, registry settings, services, etc. just sort of instantly "show up" in the target computer. (CloudVolumes puts their containers in VMDKs when used with VMs, and VHDs when used with physically-installed copies of Windows.
To be clear, CloudVolumes doesn't do any kind of isolation for "problem" apps like ThinApp or App-V, rather, their focus is on making different apps available within different VMs or while making them appear as if they've been locally-installed. They talk about wanted to make Windows application delivery, installation, and removal as simple as the way mobile apps work in iOS or Android.
Over the past few months, Harry's been referring to this as "disposable IT"—the concept that today's IT staff doesn't want to spend a lot of time troubleshooting static (i.e. "persistent") user environments, rather, they just want to hit a "refresh" button to put that environment back to a steady state and move on to better things.
This concept of course is not new. In fact it's the dream of "non-persistent" VDI that VMware's been pushing since 2006. That concept always sounded good in theory, but in practice it's been difficult to implement since all we really had was "shared base images" plus "app virtualization"—a combination which was unwieldily to manage and left a lot to be desired in the "compatibility" department. (Kevin Goodman explains it as "going through all the pain to virtualize all your apps with App-V or ThinApp just to get them to a point where you can install them on demand is like killing an ant with a sledgehammer.)
Several vendors offer solutions that are conceptually similar to CloudVolumes, including Unidesk, FSLogix, and Liquidware Labs.
Ok, so now VMware owns CloudVolumes. Cool.
I want to dig into why this makes sense and what VMware plans to do with them, but first we should take a step back and look at where VMware is on their renewed path to EUC.
First, if you've heard VMware's Desktop GM Sumit Dhawan talk in the past few months, you would have heard him talk about customers' three-phase approach to the virtualizing the Windows desktop and Windows applications:
- The first phase is taking Windows desktop applications from the physical to datacenter.
- The second phases is about disaggregation and then re-aggregation, which is much more complex since there are some real roadblocks that exist in terms of costs and tradeoffs. (Just look at all the arguments about persistent versus non-persistent images, layering, etc.)
- The third phase is about refining the user experience across devices, including things like transforming apps so they make sense on whatever form factor the user has at that moment, breaking apps into micro apps that only expose what the user needs, developing new mobile apps from existing desktop apps, etc.
Now let's map those three phases to what VMware is doing:
- VMware Horizon 6 added RDSH support, app publishing, and seamless Windows to VMware's Horizon portfolio.
- Yesterday's purchase of CloudVolumes is how VMware believes they'll be able to deliver usable images and apps that don't have the tradeoffs of Linked Clones + ThinApp.
- Stay tuned. (PowWow maybe? Others?)
Why didn't VMware just use Mirage?
I talked to Sumit Dhawan on the phone yesterday, and one of the questions I asked him was why VMware had to buy CloudVolumes. Couldn't they just use the layering that's part of Mirage?
Sumit told me that they initially looked at this, and that while the concept of the layers is similar in some ways between Mirage and CloudVolumes, the two products are actually quite different. Mirage was built for physical machines that are loosely connected which receive their layers via the network and the WAN, wheres CloudVolumes is built for always-connected clients which want to mount their layers from a fast reliable location.
I also asked Sumit about ThinApp and how it fits into this. He didn't have specifics to talk about, but its easy to infer that the ThinApp product (or at least bits of its technology) could fit nicely into what CloudVolumes could grow to be at VMware. (To be clear, ThinApp was about both isolating bad apps and providing a simple package that was easy to deploy to users. Obviously CloudVolumes handles the deployment part of what ThinApp did, but for the few apps that actually cause problems when installed next to other apps, it would be great if parts of the core ThinApp IP could be integrated into what CloudVolumes can deliver.)
What's the impact of this acquisition for the layering space?
Citing the way VMware dropped support for Citrix when they bought Desktone, Paul Stansel wrote on CitrixTips.com that he's worried VMware will make CloudVolumes a VMware-only solution. When I talked to Harry Labana on the phone yesterday, he mentioned that CloudVolumes can be very strategic to VMware and can be more than just a line-item feature in View, so it's certainly possible that Citrix XenApp and XenDesktop customers could buy it, (especially if they're running on vSphere), but overall I tend to agree that most likely this will end up being a VMware-only solution.
Paul also wrote "this may ultimately force Citrix to do is make something useful out of PVDs (Personal vDisks—technology they got from their RingCube acquisition). They promised for years that PVDs would be more flexible and roam with the users, but that promise has never been realized. PVDs are often complicated to support and one of the least mature technologies Citrix offers."
While I don't necessarily agree that PVDs are the same thing as CloudVolumes, I definitely agree that Citrix is (or certainly will now) try to position PVDs as their version of CloudVolumes. We should also keep in mind that there are other vendors who have similar solutions to CloudVolumes, (like the companies I mentioned before: FSLogix, Unidesk, and Liquidware Labs).
Actually, speaking of Liquidware Labs, I know I'm not the only one who believes this company was created for the sole purpose of being acquired by VMware, and they have to be kicking themselves again now, saying, "Seriously?? They bought those F-ing guys??" Fifty bucks says the next Liquidware product will be something to refine the Windows application user experience from mobile devices. :)
Seriously it will be interesting to see how Unidesk and Liquidware position themselves next week at VMworld as they'll both be there as sponsors. (Looks like Kevin Goodman made a fortuitous decision when he decided that FSLogix would not exhibit!)
Overall though this is a good move by VMware. Yeah, we're going to see Unidesk, FSLogix, and Liquidware talk about how their solution is better. Yeah, we're going to see Citrix try to position PVDs and/or buy one of these companies. But in the grand scheme of things, I like the concept of what these companies are doing. It fits well with my long term vision about Windows apps being around forever and many of them ultimately ending up in some datacenter with just the UIs being remoted. I might even go so far as to say that with these types of technologies, maybe something approaching a "non-persistent" disk image could work. (Not the true "non-persistent" of 2008, but something more like it.)
Congrats to VMware and CloudVolumes. Good move for both. And, frankly, congrats to Unidesk, FSLogix, and Liquidware who can now sell to 100,000 Citrix customers who will soon want this CloudVolumes-like thing too.
[Correction August 22: An earlier version of this article omitted the name of Matthieu Suiche in the list of CloudVolumes's founders.]