Hyperconvergence Performance High Score

High scores are minute, but you still can make history!

How it started

We took 12 production-ready StarWind HCAs, powered them with Intel Platinum processors, and set the hyperconverged industry performance record. Find out how we came up with this crazy idea.

Just cache-less iSCSI shared storage

Although being said to be inefficient for flash, iSCSI, like it or not, will stay here for another while. To see how efficiently all-NVMe storage can be presented over iSCSI, we disabled caching in our test cluster.

Write-back cache + iSCSI shared storage

We expected write-back caching to boost our environment performance significantly. It’s basically how we set the new HCI industry performance record! Find out how fast VMs can run with PCIe SSDs used for caching.

Linux SPDK/DPDK target + StarWind NVMe-oF Initiator

Bringing NVMe-oF support to industry-standard hypervisors is an important mission that many tried to accomplish. We used Linux SPDK target and StarWind NVMe-oF Initiator to chase for the record. Find out how efficient StarWind’s NVMe-oF implementation for Windows is.

Story ends here

Need just a quick overview of our research? We keep this article simple: All results on one page.

Buzz Lightyear said, “To infinity & beyond!” Let’s see how long StarWind’s High Score will last

Hyperconvergence Hall of Fame
1
StarWind
26 834 060
2
Microsoft
13 700 000
3
4
5