VDI provisioning is a complicated process that requires a strong understanding of the back-end hardware that hosts the desktops.
Some aspects of VDI provisioning are simple, but others require an understanding of each employee's applications, approximate logon and logoff times, and location.
Planning for CPU, which handles processing, is perhaps the simplest resource to manage because the only factor that dictates it is the number of desktops the organization needs, which is often stagnant. After IT factors in the other functions the servers must perform, VDI provisioning comes down to the number of desktops and the acceptable amount of latency an organization needs.
In certain cases, IT could cut down CPU requirements by factoring in when users log on. If all the users access their virtual desktops at the same time, the benefits of factoring in user logon time would be marginal at best.
How much memory does an organization need?
When IT professionals calculate the memory VDI requires, they must factor in user behavior and application use.
It's tempting to overestimate users' needs and invest in more memory capacity than is necessary, but this method can result in excessive hardware that could cause more harm than good. Memory is often the most expensive aspect of VDI, as well, so overspending can run up the price.
When IT pros estimate memory requirements, they should categorize workers accessing virtual desktops based on their specific needs. Kiosk workers, who only require the memory needed to run a single task, and task workers, who require only a few applications, such as email and web browsers to perform a small number of repeated tasks, require only a bit of memory.
Knowledge workers who run complex tasks and operate more applications than task workers require more memory because of their more diverse and resource-intensive application usage. The users with the highest memory requirements are power workers who use larger files and have high processing requirements because they run applications that consume a lot of resources, such as graphics editing apps. IT should categorize users based on their memory consumption and create a memory capacity plan based on each group's number of users and needs.
How should IT approach IOPS planning?
IT must ensure there are enough IOPS -- a unit of storage measurement that denotes the maximum number of reads and writes that can occur in a data transfer -- to handle all the virtual desktop storage requirements.
If IT under-provisions IOPS, users may experience desktop latency. IT should plan for more than just the average amount of IOPS that the deployment consumes.
Doing so will cover them the majority of the time, but if IOPS usage spikes during a time of high user activity, such as during a boot storm, it can leave users with significant performance delays. Instead, IT should consider peak IOPS usage as a guideline to work from for its VDI provisioning.
Run capacity tests to prevent future issues
Whether it's a wave of new users or a first-time VDI deployment, IT must often make estimates about users' resource consumption. This process can be complicated, especially for organizations that gain or lose a significant number of users.
To avoid losing money by overinvesting in virtual desktop resources and back-end hardware or hurting user productivity by underinvesting, IT should test the VDI deployment. The test should be small-scale so if something goes wrong, it only affects a few desktops. IT can track the desktop performance of users' workstations based on the VDI provisioning it allocates for them.
The test should include the same range of user types that the organization has to ensure that resource estimates are accurate. After the test, IT can determine if any desktops require more of a particular resource and apply the updated estimate to that user type throughout the organization. The test subjects can also improve VDI management by providing feedback if there are performance issues IT doesn't notice.