XtremIO / VxRail


The Virtual Desktop Infrastructure (VDI) or End User Computing (EUC) use case is really driven by two distinctly different workloads.

First, for the desktop boot image volumes, $/IOPS and latency are the key drivers for successful user and admin experiences. Scenarios, like desktop deployments, OS boot situations, login, anti-virus, application patching and other kinds of intensive IO drivers are all highly volatile and push IO requirements from underlying storage. Also, compared to server workloads, the sheer number of desktops can often lead to sudden surges in IO. If the storage infrastructure cannot scale and sustain low latency throughout these "IO Storms" it results in a very unhappy user experience. This is especially true in large-scale environments with a real-world mix of user and desktop types.

The second defining workload for VDI is general purpose NAS for user data: home directories, user profiles and some applications. These capacities tend to be very large as compared to the boot volumes and should not be overlooked as critical components to an overall VDI design. The File Consolidation workload in the Product Positioning Guide should be referenced to meet these needs.

Boot image storage is often a fraction of the overall VDI design, but its ability to deliver high performing user experience can make or break the entire solution and limit the types of desktops that may be virtualized (persistent vs. non-persistent, linked vs. full clones). These desktop options determine how much of an organization's users can benefit from a virtual desktop, how well they will work, and at what scale. For these reasons, this is the single most critical part of the design and storage must be carefully considered.

Understanding the customer's requirements and the mix of non-persistent and persistent as well as linked and full clones are all very important parts of the decision making process when considering the best storage platform for the customer's needs.

The Spiderchart below shows the distribution and weighting of the primary workload requirements for this use case.