Why is storage so important to the VDI user experience, and what are the most important metrics to measure?
In a traditional, distributed desktop/laptop environment, the storage demands of a single user laptop or PC aren't very high. Most PCs are deployed with drives spinning at 7,200 rpm or 5,400 rpm and with as much capacity as possible; 1 TB or 2 TB drives are not uncommon. These are more consumer-type devices (with consumer levels of storage reliability), where poor response time for a single user doesn’t impact the rest the desktop environment. When a standard desktop user runs a defrag job, file search or other I/O-intensive process, for example, she only affects her own work. In VDI deployments, on the other hand, poor disk response time can affect many or all users of the infrastructure, creating productivity issues that have a much greater impact. Storage that supports a VDI environment, therefore, needs to be able to deliver consistent I/O response times, irrespective of load.
In terms of determining what it will take to support the performance you need in a VDI environment, the typical Windows 7 desktop will require about 5 to 10 IOPS in normal operation. During boot time, this increases significantly -- 10-fold, to about 50 to 100 IOPS. (While these figures provide a good rule of thumb, they are only indicative and any sizing exercise should be based on an appraisal of a typical desktop in the target environment.)
When it comes to measuring VDI user experience, the most important metric is clearly performance, which can be affected directly by the response time of disk, whether opening a local file, starting up an application or powering up the desktop at boot time. Rather than measuring response in milliseconds, VDI deployments measure the number of IOPS of capacity required across all desktops in use. In addition, this workload is typically random I/O, which is more difficult to deliver with consistent performance. Capacity is less of an issue. There are opportunities to separate the boot disk from user data (which can be stored centrally, in a non-persistent VDI configuration, enabling storage savings to be made from data reduction techniques.
This was first published in June 2012