MacroSAN array takes third place in SPC-1 benchmark table

Chinese storage vendor MacroSAN has submitted an all-flash array for the SPC-1 v3 benchmark and has gained the third fastest ranking with 6,100,329 SPC-1 IOPS.

The Storage Performance Council (SPC) has a series of benchmarks and the SPC-1 v3 tests the ability of the storage array to run input-output operations, measured in calculated SPC-1 IOPS. A disclosure report shows the results summary:

Test configuration

The test configuration was made up from 8 x dual-controller MS5580G2 all-flash arrays connected by 320 x 16Gbit/s Fibre Channel links to 20 Sugon 1620-G30 servers. The controllers were in active:active setups and had 192GB cache each; 3,072GB in total.

MacroSAN MS5580G2.

The MS5580G2 arrays had 384 x 800GB SSDs and 32 x 480GB SSDs organised in;

  • 8 x 2U chassis each with 4 x 480GB SSDs and 16 x 800GB SSDs
  • 16 x 2U chassis each with 16 x 800GB SSDs.

These chassis were connected to the controllers by SAS links. The total physical capacity was 299,968GB while the used capacity (Application Storage Unit or ASU capacity) was 141,240GB. The drives were configured into 64 x RAID 10 arrays.

Top ten results 

Just one US supplier makes the top ten SPC1 v3 result. Fujitsu leads with the ETERNUS DX8900 S4 array, pumping out 10 million-plus SPC-1 IOPS. Huawei is in second place and MacroSAN’s MS5580G2 array is third. The next four places are held by Huawei systems. Korea’s Jet-Speed takes eighth spot with 2,410,271 SPC-1 IOPS.

NetApp is at number 9 with an A800 doing 2,401,171 IOPS, and that’s followed by another Huawei system.

Plotting the results  in a 2D price/performance space defined by SPC-1 IOPS and $/KIOPS we can see the Asian suppliers’ dominance clearly;

Systems to the right and low down have the best price-performance combination. We have only labelled the high-scoring IOPS results to avoid overcrowding the chart with overlapping text labels.

The IBM DS8888 result dates from November 2016. It’s not really current.

Blocks & Files looks forward to the first SPC-1 benchmark results using NVMe drives as these should go faster than SAS SSDs. Even the benchmark-winning Fujitsu ETERNUS system used SAS drives.