CHICAGO -- Storage has long been a major pain point for VDI deployments. In fact, it's one of the reasons VDI adoption hasn't taken off as seriously as people expected.
Now, advancements in storage systems may change all that. Companies such as GreenBytes Inc., Tegile Systems, and Atlantis Computing Inc. have come up with new ways to optimize storage capacity and performance in virtual environments. Here at
What were the main problems that storage brought to VDI environments in the past?
Chetan Venkatesh: There are two main challenges. The lesser challenge is capacity. When you use persistent desktops, capacity becomes an issue because these VDI VMs tend to be really large. The bigger challenge for both persistent and nonpersistent VDI is performance. That is measured in IOPS. Traditional storage architectures -- SAN and NAS -- don't have a lot of IOPS to process the kind of I/O that VDI does. VDI is write-heavy. Delivering the IOPS requires buying more storage, and that inflates the cost of VDI.
60% of all desktops by the end of this decade will be virtual.
Chetan Venkatesh, CTO, Atlantis Computing
What has changed in the landscape of storage for VDI?
Venkatesh: We can basically manufacture virtual storage by taking compute and memory and turning that into storage right there on the server. We use a form of deduplication and compression to take the RAM that's on a server and turn that into a large data store for VMs. Because we're using RAM as the media, we can solve the IOPS problem because there is a ton of performance when you use RAM as storage. With RAM, you lose all your data if you lose power, so we have a storage technology that synchronizes the contents of the RAM back to the physical storage. The storage itself doesn’t perform any I/O that is meaningful from the storage standpoint of the VM. It's a decoupling of performance from storage capacity.
What's in the future for storage in VDI and cloud environments?
Venkatesh: The biggest trend is in-memory storage. It's an important step in the next stage of storage for virtualization. The ability to take local storage means you basically change the whole convention of how storage works. The locality of data provides high performance; all data is within the same server so you can do I/O at the speed of the memory bus itself. In the longer term there are other technologies coming around -- non-volatile memory -- that will relegate the current generation of Flash storage obsolete. In five to ten years, that's what's going to happen.
How do you see VDI adoption progressing? Will it increase or be more of a niche technology?
Venkatesh: There have been a lot of expectations that the adoption of VDI would have been sooner and at a larger scale. But people don’t appreciate the complexity of what VDI has to accomplish; there were so many pieces to this VDI stack at different levels of maturity -- so it's taken many years for the right pieces to come together. Now you have a stable and scalable set of solutions that's at the right cost point.
Server virtualization was comparatively easy, because for the majority of servers, storage is not a complex challenge. But in VDI, storage is one of the most critical components, because the entire desktop experience is driven by how fast that storage system is. It's only in the last few years that companies have started to address the capacity and performance challenges of VDI.
I'm bullish on VDI now. In large enterprises, we're seeing a half to a third of all desktops being rolled out as VDI. Windows 7 migrations and adoption of Windows 8 are a large part of that. My prediction is that about 60% of all desktops by the end of this decade will be virtual.
What about Desktop as a Service?
Venkatesh: I see Desktop as a Service as very relevant in the SMB sector not in the enterprise, because of a host of reasons -- confidentiality, security, isolation -- all of those control-related challenges that enterprises care about. I think anywhere from 10-15% of adoption for SMBs will be in the cloud. Once large service providers start to provide a really high quality of service around desktops hosted in the cloud, I think you'll see significant adoption there.
Brian Madden and Gabe Knuth said in their session at BriForum that the future of Windows is middleware. Do you agree with that?
Venkatesh: Absolutely, it has already become middleware. People are not writing applications for Windows like they used to -- that is one of the biggest signs. They're writing applications for mobile devices. The existing applications that enterprises have written on Windows really are the only reason to continue using Windows as a desktop, because there is a ton of code already there that is too expensive to rewrite.