Although many companies have investigated the potential for virtual desktop infrastructure (VDI)-based environments to lower costs and leverage new work opportunities, many have discovered that VDI can be a tough technical challenge. Whereas server virtualization solutions are relatively easy to size and operate, large VDI environments introduce significant new challenges into the planning and operational stream. For example:
- VDI environments must be sized to accommodate user-visible I/O peaks that often affect many virtual desktops at the same time. Server workloads also peak, but not generally all at the same time. In the VDI world, the peaks generally occur during boot up and login times during which they are referred to as “boot storms” and “login storms”, respectively. However, I/O contention can also occur when many users perform I/O intensive actions all at once.
- Deduplication takes on new meaning. Server workloads are often very different from one another, but with virtual desktops, organizations often deploy identical desktop images. This creates new opportunity for advanced storage features, such as deduplication, to have more of an impact than ever before. These advanced features must be considered during planning.
There are untold stories regarding companies that have ventured into the VDI waters only to be scared away by the unforeseen sharks. You, however, now have an exact, shark-free map to the Treasure Island that is VDI! Tegile has partnered with Cisco and VMware to create a brand new reference architecture for building a 1,000 seat VDI environment using hardware and software provided by Tegile, Cisco, and VMware. With this exacting blueprint, you can learn how to deploy an identical environment using one of Tegile’s entry level and most inexpensive arrays to support a whopping 1,000 users.
This reference architecture document describes the architecture of the Cisco® Desktop Virtualization solution with shared storage for 1,000 virtual desktops based on VMware View 5.1 and VMware vSphere 5. The Cisco Desktop Virtualization solution includes:
- Cisco UCS® B-Series Blade Servers
- Cisco Nexus® Switching 5500 Series platform
- Tegile Zebi HA2100 storage array
This reference architecture is based on Cisco’s “scalable architecture” deployment model for desktop virtualization. In comprehensive tests while developing the reference document, Cisco, VMware, and Tegile engineers estimated that ongoing user sessions each require 7 to 10 IOPS each but that, as mentioned earlier, IOPS can peak at certain times. Engineers saw IOPS peak at around the 20,000 IOPS mark, which is well below the Tegile Zebi HA2100’s 30,000 IOPS capacity. In other words, the HA2100 can run 1,000 desktops with overhead to spare!
Test engineers used Zebi’s NFS connectivity capability as the transport mechanism of choice, although Zebi also supports both Fiber Channel and iSCSI. Zebi provided engineers with two 10 Gb Ethernet adapters on each storage controller, which are themselves configured in an Active/Active scenario. For all tests, engineers enabled Zebi’s in-line de-duplication and compression functionality to best match real-world needs.
The end result, shown in the Figure above, requires a total of eight UCS blade servers – seven servers each run 150 virtual desktop images while the eight hosts the VMware View components necessary for the environment to operate.
During testing, Zebi’s hybrid nature played extremely well. The tests resulted in a 90% cache hit rate, w which is all stored on Zebi’s solid state disks, making that access lighting fast. In almost every application test, Zebi returned response time figured below 0.05 seconds.
At Tegile, we were very pleased to see the validated results of this collaboration, proving, once again, that Tegile’s Zebi line of hybrid storage arrays are a perfect fit, no matter the workload!