This Storage Intelligence (StorInt™) dispatch covers Storage Performance Council (SPC) results . Since our last report on SPC-1, two months ago , there have been 4 new SPC-1 v3 submissions, the Inspur AS5300G5 & AS2200G2, Huawei OceanStor™ Dorado 18000 v6, and MacroSAN MS700G2-Mach. There have been no new SPC-2 submissions since our last report.
We begin our discussion with top 10 SPC-1 IOPS™ performance in Figure 1.
Figure 1 Top 10 SPC-1 IOPS
We had to increase the scale of this chart as the new #1, Huawei OceanStor Dorado 18000 v6 blew over it, with over 21M IOPS. The new Dorado system is cluster storage with 8 blades (?), each having 4 controllers and 72 1.9TB NVMe SSDs, for a total of 32 controllers and 576 NVMe SSDs. It also includes dual 100Gb (Ethernet?) RDMA links used as its inter-cluster connection fabric with 32GbFC for host access and had 32TB (4TB/blade, 1TB/controller) of cache.
Coming in at #2 was the new MacroSAN MS700G2-Mach storage system with over 11M IOPS. This MacroSAN storage was also a clustered system with 8 blades, each of which had 2 controllers and 36 1.6TB NVMe SSDs for a total of 16 controllers and 288 SSDs. The clustered storage system used 10GbE for inter-cluster communications and had 3TB (384GB/blade, 192GB/controller) of cache.
Neither of the Inspur systems reached the top 10 in IOPS, which now requires over 4.5M IOPS to place. However, the Inspur AS5300G5 system did manage well in our LRT chart shown in Figure 2.
Figure 2 Top 10 SPC-1 LRT
In LRT, the new Inspur AS5300G5 system came in at #9 with an LRT of 0.148msec. We had to reduce the scale on this chart as systems are getting so fast it’s hard to keep up. The Inspur AS5300G5 was also a clustered system with 4 blades, each with 2 controllers and 50 960GB SSDs, for a total of 8 controllers and 200 SSDs. They didn’t appear to be NVMe SSDs but we could be mistaken. No information was supplied on its inter-cluster fabric. The AS5300G5 system had 1TB (256GB/blade, 128GB/controller) of cache.
We have seen few clustered systems do well on LRT, unclear why. But Inspur’s AS5300G5 proves the exception here. Any LRT below 0.15msec seems pretty fast from our perspective.
In Figure 3, we present SPC-1’s price performance metric ($/K-IOPS).
Figure 3 SPC-1 Top 10 price-performance systems
As can be seen in Figure 3, the new Inspur AS2200G2 is our new #1 in price-performance with a cost of ~$75.76 per K-IOPS. The Inspur AS2200G2 was so cheap at $28K, we thought its pricing was a mistake. And with ~370K IOPS and an LRT of ~0.16msec, it performed exceptionally well for a dual controller system. This Inspur system had 25 400GB SSDs (we assume SAS attached) and 256GB (128GB/controller) of cache.
It’s always fun to analyze new SPC-1 submissions, which always seem to raise storage performance to new heights. With clustered systems becoming the system style du jour, it’s nice to see a new dual controller system do well on price performance. And having a clustered storage system reach into top 10 with LRT was a surprise.
As always, suggestions on how to improve any of our performance analyses are welcomed.
[This storage performance was originally sent out to our newsletter subscribers in October of 2020. If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) and we will send our current issue along with download instructions for this and other reports. Dispatches are posted to our website at least a quarter or more after they are sent to our subscribers. ]
Silverton Consulting, Inc., is a U.S.-based Storage, Strategy & Systems consulting firm offering products and services to the data storage community