Original URL: http://www.theregister.co.uk/2009/06/08/spc_1ce/
Storage body devises energy benchmark for wee arrays
But how much is it in football pitches?
The Storage Performance Council has produced a storage energy benchmark, which will allow users to compare the energy consumption of storage arrays against a recognised standard.
It is the SPC-1C/E, where the E stands for energy. This is an extension of the existing SPC-1C benchmark, based in turn on SPC-1. IBM and Seagate results are available on the SPC website. The benchmark only includes small storage devices with up to 48 drives - disk or solid state (SSD) - in an enclosure or enclosures up to 4U deep overall and some kind of HBA or controller head.
A future SPC-2C/E benchmark will cover larger storage configurations. This will use one of the three existing SPC-2C workloads in its performance phase: large file processing, large database query, or video on demand.
The SPC notes that energy consumption measurements are taken during both idle and active states of the SPC-1C/Ebenchmark execution. Multiple idle modes are allowed and the benchmark highlights anticipated energy use in environments that impose zero (idle), light, moderate, or heavy workload demands upon the benchmark storage configuration.
The active (performance) state consists of SPC-1C-equivalent performance test runs, and SPC-1C/E energy use results cannot be reported without these test run results. A more comprehensive description of the SPC-1C./E benchmark can be seen here.
Benchmark results are reported in some detail and enable average annual energy consumption and cost figures to be presented. The SPC may introduce an end-user tool to enable the results to be made relative to a customer's local conditions. This could involve allowing the customer to input their data in the hours-per-day value in the benchmark calculation and using local energy costings, for example.
Seagate announced SPC-1C/E results for a 24-drive array, configured with around 7TB of storage, using Savvio 10K.3 2.5-inch hard drives. It reported an Annual Energy Use of 1,765.41 kWh and a projected Annual Energy Cost of $211.85 at $0.12/kWh, with a rating of 8,013.39 SPC-1C IOPS.
IBM produced SPC-1C/E figures for its System Storage EXP12S product. This is a 2U system configured with eight 69GB SSDs in the benchmark. The Annual Energy Use was 1,425.41 kWh and the projected Annual Energy Cost was $171.05.
Interpreting these results is not simple. There are composite metrics for nominal power, IOPS, and IOPS/Watt, which are calculated from hours per day of heavy and moderate use and idleness. These are further modified by low, medium, and high daily usage. The picture is a complicated one and it seems unlikely that the storage industry will be able to devise a storage energy figure as simple (relatively) to interpret as a car's fuel consumption.
Knowing the energy use and energy cost is relatively useless without knowing the work the array has achieved and how much storage capacity is involved.
In the IBM case, the SPC-1C result was 45,000.20 SPC-1C IOPS from a near 560GB array. The benchmark executive summary report does not present an energy use per SVP-1C IOPS number.
If it did then the IBM result would be 0.0317 kWh per SPC-1C IOPS. The Seagate one would be 2.170 kWh per SPC-1C IOPS from a 7TB array. This shows that the SSDs were far more energy-efficient than the Savvio hard drives, but were based on a much smaller storage capacity pool.
EMC has not supported SPC benchmarking, so we shouldn't expect EMC SPC-1C/E numbers to become available. ®