For all the benefits desktop virtualization offers for certain applications, there is an ongoing discussion around the back end storage requirements — and the potential costs. Ruben Spruijt penned a lengthy tome about it back in December, and with Win 7, we can be sure that file and applications sizes are not going to be getting any smaller.

Storage Performance Requirements with VDI
There are some slick demos out there these days that demonstrate how much better user experience can be for rich media applications over VDI. But what you don’t see is the price tag associated with the storage and server infrastructure required to support this kind of performance. Here are some things to consider:

• As Ruben points out, because VDI requires a synchronous connection with storage, the quality of the storage network is going to have a direct impact on the quality of the user experience. High capacity, high performance drives = big $$$.
• A good storage network isn’t enough. With VDI, user data is stored separately than the base operating system and applications, so you’ll need some beefy servers to prevent lags there.
• The actual storage network can become the bottleneck in larger deployments, so you should make sure the SAN and Fibre Channel infrastructure can handle the traffic. For a 1000-user deployment, you’re going to need 20-40 Gbit of network connectivity to storage, with multiplexing and other costly features.

Yikes. No wonder the storage companies are excited about VDI.

Storage Performance Requirements with Wanova
With Wanova, you can get the same centralized management and data protection, but at a dramatically lower cost. Each endpoint runs a cache of the centralized desktop, leveraging local (and cheap) compute power and storage. Consequently, we don’t need an expensive server farm to handle execution. Plus, you get great side benefits,  such as the ability to work offline. Since our architecture is distributed and asynchronous, we don’t need Tier 1 storage — you can use Tier 2 or Nearline storage that is 1/5 the cost.

To summarize just some of the methods Wanova uses to ease storage requirements:

• Wanova leverages the cheap endpoint disk for executing the Windows and application workloads locally. Storage cost for workload execution on the client is $0.2/GB vs. $5/GB on the server side in VDI. And that’s excluding server side networking, NAS/SAN equipment, etc.
• With Wanova, the endpoint acts as a ‘shock-absorber’ to dampen any IO bursts and load during execution, for example, filtering page file access, temporary file access, log and repeated file read/writes, etc. This reduces the load on the Wanova Mirage server to handling only persistent changes on the desktops. So, with Wanova, the server and storage subsystem must only be sufficient to handle the averaged and deduped volume of changes.  It is interesting to see Daniel Feller’s comment to Ruben’s article, highlighting the issues with sizing a VDI deployment using average IOPS. These issues do not exist with Wanova.
• Another interesting fact raised in Ruben’s article is that with VDI, the typical IO workload is write heavy, with ratios mentioned in the order of 80:20 writes to reads. Write throughput is much more expensive than reads in RAID configurations, as Ruben explains. With Wanova, not only is less IO required at the server side, but the write:read ratio is actually reversed, allowing cheaper Nearline storage to be used with fewer, larger capacity disks — further enhancing the cost efficiency of the solution.

We recognize that there are applications where VDI makes great sense. However, for organizations that want to centralize control over their roaming force of laptop users (while still keeping them happy with good performance, personalization, and offline use), Wanova offers a far more cost-effective solution.