April 13th, 2017 by Kevin OBrien
VMware Continues to Push the HCI Market with vSAN 6.6
VMware just launched version 6.6 of vSAN, bringing another round of new features and enhancements. With more than 7,000 customers, and growing rapidly, the sixth version of vSAN boosts performance of its all-flash configurations upwards of 50% in some situations. It also supports Intel's latest Optane flash technology at launch, further driving down latencies for demanding workloads. As vSAN hits version 6.6, I'm continuously reminded as to the fast pace development cycle brought by VMware, strengthening their leadership position in HCI.
We've been quite impressed with the performance and feature enhancements over past iterations of vSAN. When we first looked at it as version v5.5, the main focus was cost-effective performance through use of a flash tier above hard drives for capacity. v6.0 brought support for all-flash configurations as well, further driving down the cost model with data-reduction capabilities, was an update in v6.2. VMware has added a number of key features over this time, including some of the following I really liked:
- 2-Node ROBO
- UltraDIMM and NVMe Support
- IOPS limits through QOS rules
- iSCSI Support
- All-Flash licensing supported in all versions
With the latest vSAN 6.6, VMware has brought more value to the mix highlighted by:
- Native Encryption for data-at-rest
- Day 0 support for Intel Optane (useful for low-latency sensitive applications)
- Up to 50% boost in IOPS for all-flash configs with optimized checksum and dedupe
Encryption is an important tool for long-term storage platforms, especially as arrays near the point in time of being decommissioned (or even transported to new sites). Preventing unauthorized access to data through encryption keeps customers happy and legal battles fewer. With storage performance being one of my main focuses though, seeing support at the onset for new technologies is always nice. This is the case with Intel's NVMe-based Optane, which was supported by VMware right out of the gate. Lastly, as VMware continues to refine and improve the vSAN codebase, they were able to eek out "up to 50%" greater IOPS in all-flash configurations.
As readers recall from our vSAN All-Flash review, we spent a lot of time showing overhead differences in vSAN between raw and data-reduction services being enabled. Hearing that performance will jump in non-inconsequential increments is always a good thing, which we hope to test with our own workloads soon. Talking with the vSAN team, the 50% increase in performance comes from a synthetic workload measuring 4K 70%R/30%W performance. Where the prior tested configuration of vSAN topped out at 100,000 IOPS, the new codebase pushes that number to around 150,000 IOPS. We're very curious how that changes results in application workloads.
Overall the continued improvement and value-adds VMware continues to bring to the vSAN product has been fantastic. From the start the team hasn't been shy on talking about performance, which contrasts starkly with many HCI vendors that dread the topic. While we have yet to deploy a vSAN 6.6 cluster in the lab, I am anticipating to log hours on the platform. For now, Duncan Epping has a deep overview of what vSAN v6.6 has to offer.