By now, it’s no secret that organizations considering moving to a virtual desktop infrastructure (VDI) need to plan carefully around the storage that will support the infrastructure.
According to a recent study conducted by VIBriefing and sponsored by Virsto Software, 46% of nearly 500 IT professionals surveyed said their VDI projects had stalled. The primary reasons for the delays were unacceptable user cost overruns (31%), inadequate user performance (29%), software licensing costs (22%) and prohibitive storage price requirements (19%). As a prominent cause for poor user performance, storage is probably the culprit for close to half of those stalled projects.
VDI storage benchmark and planning tools can help by giving administrators a better idea of their storage requirements before beginning a project. These VDI benchmarking and planning tools can alleviate surprises that lead to cost overruns and poor performances by telling organizations if their existing storage system or one they’re considering buying can support a real VDI workload.
VMware Inc. publishes a reference architecture for shops considering VMware View to help determine what kind of IT infrastructure, including storage systems, they would need to support a particular VDI workload. VMware also offers professional services to help with the project assessments, including storage benchmarking.
“At the end of the day, the storage sizing, whether it’s with NetApp or EMC or Dell or HP or others … [is] a shared SAN or NAS infrastructure. That bucket of storage, since it’s shared by design, [is] very difficult to size because it’s about performance versus capacity,” said Mac Binesh, group product manager for enterprise user computing at VMware.
IT organizations can use VMware View Planner to test a storage system and IT infrastructure for its ability to handle virtual desktops. View Planner simulates user workloads for different types of users. “We measure performance of [an] application, open times, close times, and we also support executing many types of applications and many variations of the applications,” Binesh said.
Citrix Systems Inc. also provides an architecture guide for equipment sizing for XenDesktop. And Citrix recommends Login VSI from Login Consultants for customers interested in generating a workload to test their environment’s ability to handle VDI.
“Login VSI mimics a lot of the daily desktop operations -- opening and closing files and launching applications,” said Kevin Strohmeyer, director of product marketing, enterprise desktops and apps at Citrix. “And there are standards where you measure, for instance, how long [it takes] to launch an application or open up a document. Those are some of the actions that tend to be IOPS-consuming from the storage side.”
In addition, Citrix offers the Desktop Transformation Accelerator, a tool that provides step-by-step guidance for VDI assessment and design. “It records information about what [a user’s] project is trying to do and wherever they are in that planning project, it provides the right tools that are needed,” Strohmeyer said. “There’s a whitepaper and architecture guide for storage and VDI, and then there’s a calculator that they can fill out that would provide [capacity and IOPS requirements] to them.”
VMware View Planner and Login VSI require an actual system buildout to test how well an existing or new IT infrastructure will be able to support VDI. But they provide information about the entire infrastructure -- the CPU, the memory, the network and the storage.
For IT shops looking to gauge how well a storage system would support VMware View, the VDI-IOmark tool from VDI-IOmark.org can run a pre-recorded simulation of a VDI workload. The tool was designed to reduce the cost and time from the assessment of a storage system’s suitability for a planned VDI implementation.
VDI-IOmark was created by IT analyst firm Evaluator Group in partnership with storage vendors.
“We set up a real VDI environment. We ran real VDI user workloads. And then we recorded each and every I/O operation that was made by that system,” said Russ Fellows, senior partner at Evaluator Group.
When a user runs VDI-IOmark, the recording is replayed on a single server connected to a storage system. “To the storage system … it looks exactly the same [as a live workload],” Fellows said.
VDI-IOmark’s test configuration allows for 64 unique “streams,” or recorded VDI sessions. Each VDI session runs the same workload, a combination of seven applications (Microsoft Word, Excel, Outlook, PowerPoint, Internet Explorer and Windows Media Player; Adobe Acrobat; and 7zip, an open source archiving utility). The first 64 streams can be scaled to match the number of users an IT shop plans to include in its VDI project.
“Each [stream] is essentially the same, but things happen in different sequences, different orders. Some generate more I/O than others,” Fellows said. “The first 64 users are unique. Beyond that they’re copies, but we time-shift them, so each group of 64 gets shifted out by, say, 90 seconds or some amount. We’ve run thousand-user tests before, and there’s really no limit.”
VDI-IOmark also is planning a light-user profile for the storage benchmark, which could be applied to all or a portion of streams in a workload test. A light-user stream would run only some of the seven applications.
Fellows said that a VDI-IOmark test can be run in an hour or less and is more efficient than VMware View Planner or Login VSI in terms of the hardware required to run it. “We can run well over a thousand VDI users on one physical server, and if that physical server is super-beefy, like 24 or 32 CPU cores and 256 GB of memory, you can support maybe 3,000 or even 4,000 users on that one system,” he said. “It would take literally 10 times that number of servers [with a tool other than VDI-IOmark].”