Learn why Intel® Optane™ SSDs are a better option than NAND Flash SSDs in the storage cache tier.

When configuring a new VMware vSphere and vSAN cluster, one common goal is to match the workload requirements to the characteristics of the hardware components in the host configuration, with the intention of optimizing cost and performance and achieving optimal service levels.

All host components influence this sizing process, but sizing NAND Flash SSDs can be more challenging than other components because the I/O capabilities—MB/s and IOPS—can change with workload profiles. For instance, the read/write throughput and read latency of a NAND Flash SSD varies as concurrent write I/O activity increases. These differences should be considered when sizing a storage solution, especially when sizing the cache tier.

Drive endurance is another factor that should not be overlooked when sizing a cache drive. If we simply follow guidelines for minimum endurance levels in terabytes written (TBW), without matching the workloads to the drive endurance, we risk using drives that can wear out before the warranty lifetime. Although the supplier may replace the faulty drive, the failure can still expose the business to data loss risk and service-level effects. For instance, in a VMware vSAN cluster, if a cache drive fails, the entire disk group becomes unavailable—increasing the data loss risk until the entire disk group is replaced.

In this white paper, we compare the differences in VMware vSAN performance using a NAND Flash SSD versus an Intel® Optane™ SSD. We evaluate how the NAND Flash SSD performance capabilities vary with different read/write I/O mixes. That is to say, the gradual increase in cache-to-disk de-staging rates can affect overall performance; this explains why the Intel® Optane™ media can sustain higher throughput than a NAND Flash SSD when processing the same mixed read/write load.

We look at the different endurance capabilities and draw attention to the growing risks of data loss if a drive is selected using outdated endurance guidelines. And finally, we summarize our findings, establishing a comparison of cost, performance, footprint, and risk of data loss illustrated by two cluster configurations.

Generic performance and endurance guidelines are commonly used to select cache drives. These guidelines help prevent the selection of inadequate drives, but when they are used as recommendations for which drive to use, the opportunity to optimize for cost and performance can be missed. In some cases, using a drive that does not provide acceptable endurance for certain workloads can increase exposure to data loss.

In this paper we provide information that is not included in the generic sizing guidelines. We describe how Intel® Optane™ technology in the cache tier adds value to a multi-tier distributed storage implementation like VMware vSAN. This value results from the Intel® Optane™ SSDs higher throughput and endurance, compared to NAND Flash SSDs:

• When processing a 50 percent read/write load, Intel® Optane™ SSDs can provide up to 2.5X more throughput.
• The Intel® Optane™ SSDs lifetime endurance is 16X more than a NAND Flash SSD.