In considering use cases where PMem plays a relevant role, Virtual Desktop deployment is one of the top solutions. Ideal for memory intensive scenarios, PMem introduces a considerable cost saving factor, as one of the main design goals for a Virtual Desktop Infrastructure (VDI) architecture is to maximize the number of desktops per ESXi host, while providing an excellent user experience.

In a large VDI system with 4,875 virtual desktops, 4,000 of them powered on, deployed across 41 Dell PowerEdge R740 hosts running vSphere and vSAN, where the average consumption per VM is as follows:

• 4.35 vCPUs/ 8 vCPUs per core
• 10 GB consumed memory/ 1 GB active memory
• 140 GB usable storage

Identical deployments were compared, with one using DRAM only and the second using about 25 percent of the DRAM amount combined with PMem. The results are summarized in the following chart:

Figure 14.    VDI Use case DRAM vs DRAM and PMem

The DRAM and PMem configuration is proven to be 15 percent less expensive than the pure DRAM one. As stated in the best practices section, when active memory is below 25 percent of the consumed memory (and in this case, active memory is 10 percent of consumed memory 1 GB to 10 GBs), PMem plays a significant cost savings role.

A VDI solution providing more memory to each user could result in improved user experience and more productivity. Supporting more users per server also reduces the IT infrastructure that is required to host a VDI solution.

On the performance front, two generations of Dell PowerEdge servers were tested: The PowerEdge R730xd and PowerEdge 740xd, the PowerEdge 740xd with Intel Optane PMem. The solution with Intel Optane technology enabled us to support 3.75 times as many users as a legacy solution without Intel Optane memory.

Note: More information about this performance comparison is available in the PowerEdge R740xd with Intel Optane DC persistent memory document here.