Adopting virtual desktops is a project that sounds appealing in theory, but can be quite tough to pull off in reality. Performance is one of the greatest roadblocks deterring VDI adoption, and more often than not performance issues can be directly traced back to storage infrastructure. Laptops these days are flash-based and users have become extremely comfortable with the prompt nature of flash, leading them to expect instant responses from virtual desktops too.
In order to lower the cost for each virtual desktop, the largest possible number of virtual machines need to be loaded onto the least possible physical servers which are connected to a single storage array. Usually CPUs on physical machines have ample power to support a large number of virtual desktops. However, a high concentration of virtual desktops leads to storage I/O conflicts. Even though the IOPs requirement for each virtual desktop is low, having hundreds of desktops rely on a single storage system can lead to bottleneck issues.
Finally, providing consistent and reliable performance is no mean task. The storage system has to deal with login storms at the beginning of the day, anti-virus scans, updates, and other I/O requests that employees generally create on a typical day.
VDI performance challenges can be tackled while keeping costs low. Most of the solutions to handle VDI performance challenges are based on the use of flash storage, since hard disks cannot keep pace with the enormous random I/O needs of virtual desktops.
- All-flash: All-flash arrays are unrivalled when it comes to performance. While cost is certainly a challenge with all-flash arrays, the upside is that they can cater to a substantially higher number of virtual desktops and deliver performance on par with physical desktops.
- Hybrid-flash: Since flash storage costs have plunged downward, hybrid arrays these days come with about 25% flash, reducing the possibilities of a cache miss. Hybrid flash arrays also comprise of a HDD layer that can be used to exclusively store user data. In this structure, flash tier storage is allocated to the desktop and its applications.
- Server-side flash to reduce latency: Server-side flash can act like a Band-Aid for VDI projects that use HDD storage. Server-side flash products normally use a combination of server flash storage and server-based caching solutions. Most server-based caching solutions these days can cache both read and write I/O.
- Data efficiency: Majority of all-flash arrays and a few hybrid arrays come with in-built efficiency techniques such as deduplication, compression, and thin-provisioning to enable the storage array to minimize storage consumption without decreasing VDI performance. Offloading these tasks from the VDI software to the storage system helps the processing power of the physical host.
Providing consistent and reliable storage performance for VDI while being budget friendly is definitely a grim challenge. The good news is that with the proliferation of flash, there are multiple options available to address VDI performance issues. One of the most effective ways to keep prices low is loading as many virtual machines as possible on every host while utilizing flash to resolve storage I/O contention.
Latest posts by Anupam Bhide (see all)
- Is the Public Cloud in decline? - May 29, 2019
- Will Blockchain disrupt the world of data storage as we know it? - April 25, 2019
- The Problem with Enterprise Data Storage and One Promising Solution - March 25, 2019