Many storage performance issues can arise when implementing a virtual desktop infrastructure (VDI). VDI projects have difficult workloads, and not having an adequate storage system design or architecture can lead to cost/performance rates that greatly exceed your return on investment (ROI) expectations.
In this interview, Russ Fellows, senior partner at Evaluator Group, details the VDI storage requirements VARs need to consider, common storage-related problems in a VDI project and how to avoid them, and other ways to best support storage in a virtualized desktop environment. Read the transcript below or listen to the podcast.
Download for later:
- Internet Explorer: Right Click > Save Target As
- Firefox: Right Click > Save Link As
SearchStorageChannel.com: First, why is storage so important in a VDI environment?
Fellows: We’ve been speaking with a lot of end-user IT administrators and architects, and we have a large base of end-user customers. We first started talking to people a couple years ago that were coming to us with performance issues when they were implementing VDI, and it always came back to storage. They were seeking some input from us, so we did quite a bit of research and what we found is that VDI can just be a very difficult workload. Just throwing money into your storage system hoping to solve problems the old-fashioned way, without applying a good architecture or design, can lead to cost/performance levels that far exceed any ROI that people might use to justify VDI. VDI is usually justified on a dollar-per-user basis, so you really have to apply a lot of intelligence and it almost always comes back to the storage system.
SearchStorageChannel.com: How would a VAR or systems integrator go about starting to research and plan for storage requirements for a new VDI deployment?
Fellows: It starts with the architecture and the design of the VDI deployment. You have to look first at whether you’re planning to use pooled [desktops] (also known as linked clones) or persistent [desktops]”, which [are] more like traditional desktops. So that’s one of the first decisions, and then from there you can start designing your storage system and picking out the features that are going to be required to give you the best price/performance.
SearchStorageChannel.com: And what are some common, storage-related challenges that can arise in a VDI project?
Fellows: Again, it comes back to storage and storage performance issues. And what happens a lot of times is people will do a proof of concept in larger environments, but they don’t necessarily test for performance. So they test for features and functionality to make sure that things work, and may even test it in the environment that they plan to use it, but only at, you know, one-quarter or one-tenth of the planned number of users. And then when they put the full load on it, then the system falls over because in a lot of cases the storage system just can’t handle the I/O workload. So, just using the back-of-envelope calculations that some of the VDI vendors publish in their guides -- sizing 8 I/Os per user or 10 I/Os per user -- those are almost completely irrelevant, and you really need to do a lot more investigation and testing.
SearchStorageChannel.com: Can you share some tools or best practices to help avoid some of those problems?
Fellows: [There are] a few tools that we’re aware of. VMware has two tools: One is called RAWC, [and] another is called View Planner. There’s another one called Login VSI, and then we worked with some industry vendors and end users and developed another tool called VDI-IOmark, which can run a VDI-specific workload for storage systems.
SearchStorageChannel.com: For a VAR whose customer has already implemented VDI and discovered that the storage is inadequate, how do they go about fixing the problem?
Fellows: That’s something that we’ve seen happen quite a few times. And [there are] two ways that you can address that, and they’re not mutually exclusive; in other words, you may want to do both. One is that you can look at using some tools that can add some intelligent mapping of VDI to storage systems and make the storage system a lot more efficient and more effectively use high-performance solid-state storage (SSD). Some of those tools have a variety of names, like layering software, but Wanova and Unidesk are two great examples. In fact, we have Unidesk in testing in our lab right now; I’m just doing an evaluation of that product.
And then there’s some other appliances out there -- caching appliances, storage systems -- that are designed to solve this issue, and they go in between the storage and the VDI compute system; software appliances like Atlantis Computing or Virsto are some examples there. On the storage side there’s a whole bunch of startups that are looking to address VDI, companies like Nimble [Storage] and Tintri and even all-SSD systems like Nimbus. So ultimately it’s important that a storage system has to contain some amount of high performance, usually SSD, as a tier. It can also be used as a cache, but if you do this with an intelligent architecture and have the right storage systems features you can really get excellent price/performance with VDI.
This was first published in March 2012