BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
What hardware or software can be used to address VDI performance issues such as boot storms and login storms, and what role does storage tiering play?
Boot storms and login storms are examples of high I/O demand created when large numbers of virtual desktops are rebooted or during periods or high login or logout activity, typically at the start or end of the user day. The bulk of the increased load is read activity, and so hardware that provides additional caching can help mitigate these VDI performance issues. This hardware includes cache cards such as NetApp’s Flash Cache (formerly Performance Acceleration Module, or PAM) or solid-state drives (SSDs). There are also software solutions, such as Altantis Computing’s ILIO, that enable the underlying hardware to serve I/O more efficiently.
Where linked clones or deduplication has been used, the clone master image should be placed on the fastest storage device (for example, SSD) as most read requests will be targeted at this data. Where devices such as NetApp Flash Cache cards are used, their use should be restricted to common OS data and exclude user data (which can pollute the cache and reduce the overall performance improvement).
Storage tiering can play a vital part in delivering a cost-effective VDI solution. Not all the data within a desktop needs fast response. Tiering provides the opportunity to deliver faster responses to parts of the virtual desktop that need to be loaded quickly, for instance, at boot time. Where deduplicated storage has been used to store virtual machines (e.g., FlexClones or linked clones), there will be certain portions of shared data that will be accessed concurrently by many virtual desktops. This data should be placed on the fastest storage device, either directly or using the dynamic tiering options of today’s storage arrays.
Dig Deeper on Flash storage for virtual workloads
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.