Expanding memory capacity is key to dealing with the rising tide of worldwide data growth, fueled by trends such as high-definition video surveillance, the IoT, AI, and 5G connectivity. This data inﬂation places a heavy emphasis on memory expectations. IT knows that the virtual machines (VMs) tasked with handling these workloads will experience periodic ﬂuctuations and occasional peaks. Having insufficient memory will yield disk swaps that lead to performance bottlenecks, so IT often pursues a strategy of expanding memory capacity for peak loads, even though true utilization rarely reaches those levels. However, provisioning more memory per VM than needed reduces the total number of VMs possible on a given server. This impairs total resource utilization (due to lower VM density) and infrastructure ROI. Unfortunately, IT staﬀ often don’t get to see the larger economic picture. They’re doing their jobs, working to make sure that applications perform quickly and reliably. But they, and even executive management, may not grasp the true costs of provisioning more VM memory with DRAM than strictly needed.
A 2018 report from IT management solution provider ControlUp details a survey including 71,845 server VMs spanning over 900 organizations worldwide. The survey found that 77 percent of servers were over-provisioned. Moreover, across all servers examined, any given server was over-provisioned by an average of 55 percent. All told, ControlUp pegged the amount overspent on this unneeded DRAM at USD 23.8 million. Additionally, over-commitment of DRAM leaves less memory available for adding more VMs, resulting in lower CPU utilization per physical server, and thus higher TCO.
Organizations need a way to achieve higher server memory capacities—and the high VM performance that can come from having right-sized memory allocation—without the potentially steep per-gigabyte costs of DRAM. IT can future-proof data centers by adding persistent memory—specifically, Intel® Optane™ DC persistent memory, which provides a path to aﬀordably expanding system memory while maintaining the high-performance levels associated with running large in-memory workloads. Intel® Optane™ DC persistent memory modules (DCPMMs) ft in the same DDR4 slots as conventional DRAM, making them easy to install while delivering capacities several times that of most DDR4 modules. Businesses can keep running DRAM for their most performance-sensitive “active” data and deploy Intel® Optane™ DC persistent memory to satisfy remaining memory needs, such as occasional peak loads or high seasonal cycles.
By having more total system memory available, managers can make sure application owners retain all the memory they expect, which in turn will help save considerable time on handling resource requests.
With this new technology, organizations can achieve up to a 25 percent increase in VM density while saving up to 17 percent in hardware costs per virtual desktop infrastructure (VDI) session. Alternatively, administrators could maintain the same number of VMs per server while reducing total server cost. Intel® Optane™ technology opens multiple routes for improving efficiency and boosting TCO savings. In turn, those substantial savings can allow your budget to drive innovation and company-wide growth.