Virtual Desktop Infrastructure (VDI) provides a way to manage user desktop environments with fewer resources, increased manageability, enhanced security, and greater performance. With VDI, a single image can be used to propagate thousands of desktops. Because this reusable image is built in accordance with company security standards and an approved software stack, it provides IT professionals with a tremendous level of control over their virtualized computing environment, making it far more secure and reliable.
VMware Horizon 7 lets you create, deploy, and update desktop images. It is a centralized desktop virtualization solution that helps you deliver virtualized desktop services and applications to end users from centralized VMware vSphere servers.
The tests described in this paper demonstrate the ability of the Dell EMC PowerMax storage system to support VDI deployment with VMware vSphere and Horizon 7. To perform the tests, we used LoginVSI, the industry standard for load-testing VDI solutions, to generate load and analyze the system’s capabilities. In addition, we collected and reviewed storage performance metrics and data reduction efficiencies.
The test environment hardware consisted of a single-brick PowerMax 8000 storage array and twelve Dell PowerEdge R740 servers as compute nodes for desktop deployment. Other servers were used for management and monitoring. We used this environment to generate and support 400 full-clone desktops, or 600 instant-clone desktops running the LoginVSI “Knowledge Worker” workload, which is a well-balanced user type. LoginVSI output showed an exceptionally good base performance (“VSIbase”) and a very high estimated maximum performance (“VSImax”). It also showed that the environment could have scaled to support more than twice the number of compute nodes, demonstrating the ability of PowerMax to accommodate intensive workloads.
Before LoginVSI workload phase begins, all the desktops are restarted, creating a ‘boot-storm’. Boot-storm is an important aspect of VDI deployment testing, demonstrating what might happen if all the desktops need to start simultaneously after planned or unplanned downtime.
The tests demonstrated high IOPS (over 110,000), high bandwidth (near 6 GB/s), and low latencies (1.4 ms read response time and 1.8 ms write response time) during the boot-storm phase with full-clone desktops. The tests demonstrated even lower latencies (0.4 ms read response time and 0.2 ms write response time) during the LoginVSI workload phase with full-clone desktops. In comparison, instant clones demonstrated sub-millisecond latencies during both the boot-storm and workload phases, because instant clones in general are more efficient.
Instant clones share a virtual disk and therefore provide their own inherent deduplication value; however, PowerMax compression still provides strong data reduction benefits. With full clones, each desktop gets their own virtual disk. In this case, PowerMax compression and deduplication values are both of great benefit for data reduction.
While deploying 400 full-clone desktops, PowerMax provided a data reduction savings of 97% per desktop by reducing the average desktop storage size from 13 GB raw capacity to 0.4 GB actual allocated capacity (12.6 GB saved capacity / 13 GB raw capacity = 97%). These savings were noticed immediately after the desktop initial deployment phase. After completing the LoginVSI workload phase, the savings were 94% per desktop as the average desktop storage size was reduced from 13 GB raw capacity to 0.84 GB actual allocated capacity (12.16 GB saved capacity / 13 GB raw capacity = 94%). Capacity saving benefits of the workload phase will vary with the amount of data that is changed.
These tests prove that PowerMax is an excellent storage infrastructure choice for deploying the VMware Horizon VDI environment, maintaining high performance, low latencies, and strong data reduction efficiency.