Reg comments16

DataCore’s benchmarks for SANsymphony-V hit a record high note

65.7% of the performance of a NetApp box for 2.9% of the cost

+Comment DataCore's SANsymphony-V software with a parallel IO feature has produced a radical advance in price/performance on the SPC-1 benchmark.

SPC-1 is an objective test and measures the data access performance of a storage subsystem. It enables meaningful performance comparisons to be made between different vendors' systems.

The DataCore Lenovo server system produced 450,290.87 SPC-1 IOPS, the 9th best result. The leading system is a Hitachi VSP G100 flash-accelerated system that produced 1,004,941.89 IOPS. The 5th best system is a NetApp FAS8080EX all-flash one that produced 685,281.71 IOPS with a price/performance of $2.77 in April 2015. Its price was $1.898 million.

The DataCore/Lenovo system has a price/performance figure of $0.08, the system costing $38,400 and 29 cents. We understand this is an SPC-1 price/performance record.

Think on that for a moment.

DataCore_SPC_1

SPC-I IOPS ranking chart

The DataCore/Lenovo system delivered 65.7 per cent of the performance of a NetApp FAS8080EX for 2.9 per cent of the cost. Put it another way; the NetApp array costs 97 per cent more in price/performance terms. Tell me that isn't amazing.

Some more good news. The DataCore/Lenovo system had an average response of 0.32 milliseconds at 100 per cent load. We think this is an SPC-1 record as well.

The SPC-1 IOPS chart-topping Hitachi VSP G1000 had a 1.15 millisecond average response at 100 per cent load, 72 per cent slower and costing 98 per cent more.

This is just unreal.

DataCore_SPC_ Price_Performance

What we can certainly say is that DataCore's software uses cheap commodity server and SSD hardware to blindingly good effect.

Hardware configuration

The DataCore configuration has DataCore's SANsymphony-V v10.0 virtual SAN software running in a Lenovo X3650 M5 server with 2 x 14-core Xeon E5-2695 v3 processors and 544GB of main memory. The operating system was Windows 2008 R2 Enterprise Server w/SP1.

The configured storage capacity was 10.1TB and it was operated in a RAID 1 mirrored fashion. It was composed from 16 x 480GB Samsung SM863 SSDS hooked up via a 6Gbit/s SATA interface. There was a single 500GB Samsung 850 EVO SSD linked to an internal controller for page swapping.

This server had a single internal 300GB 10K SAS disk drive for system software, plus 8 x 300GB 12Gbit/s 15K SAS disk drives – HGST Ultrastar C15K600 mounted externally.

Reg Comment

Why did the DataCore system perform so well? It has an Adaptive Parallel IO feature in SANsymphony-V v10.00, which has the server processor's cores running IO in parallel instead of queuing them up through a single core.

It appears that DataCore's parallel IO virtual SAN running a server with SAS-connected flash storage is the 9th fastest SAN in history in SPC-1 terms, the runaway price-performance leader and the one with the lowest latency at 100 per cent load. We cannot see any way in which an externally-connected SAN, using Fibre Channel or Ethernet, could be faster, because parallel host server IO is so fast.

We would love to know how the DataCore software performs in a server using PCIe-connected flash, as this would cut out the SAS protocol stack and should produce an even higher SPC-1 IOPS number.

Get the full disclosure DataCore SPC-1 report here [88-page PDF]. ®


Biting the hand that feeds IT © 1998–2017