One month after completing its first-ever benchmark tests on storage systems, the Storage Performance Council has released its full disclosure reports on
The SPC-1 benchmark tests certain aspects of a given storage configuration, including the maximum input/output (I/O), the total storage capacity read and written during the course of executing the benchmark, the data protection level and the average system response time.
IBM's Enterprise Storage Server F20 with a total storage capacity of 1,201.49 G Bytes and running a RAID 5 configuration earned 8,009.44 IOPS (I/O per second) and a response time of 2.99ms, The total cost of the system with a three-year warranty was $357,100.00.
Sun Microsystems' tested a 343.51G Byte, mirrored Sun StorEdge 9910, which garnered 8404.22 IOPS and a response time of 2.07ms. The StorEdge 9910 cost $624,377.44.
The third system -- and the fastest -- on the list was LSI Logic Storage Systems' E4600 Fibre Channel Storage System. The E4600 featured a mirrored configuration with 400G Bytes of storage capacity. The E4600 scored 15,708.17 IOPS and a response time of 1.64 ms for a cost of $251,434.15.
As the results show, the smaller LSI system bested the monolithic arrays from Big Blue and Sun, but what does that say about the SPC test in general?
The SPC claims the new benchmark specification is the first vendor-neutral process to accurately compare and configure direct-attached or network storage technology. The overall goal of the SPC Benchmark 1 (SPC-1) is to give end-users and integrators benchmark results for every enterprise storage system in the market, but questions have been raised by analysts and vendors as to the real value of the SPC's numbers.
Christian Ober, principal storage analyst at Ideas International Ltd., Sydney, Australia, said the SPC-1 benchmark is designed to emulate multi-user I/O applications such as database/OLTP and mail applications.
Ober said that the E4600 was the only product in the group that utilized a modular architecture, which shows what price and performance gains can be achieved versus monolithic architectures of the other two test products.
An argument can be made that the SPC-1 testing is subjective at best. The three storage systems that underwent testing had different architectures, RAID levels and capacities. There were also variations in platform support and software and management features, all of which can impact performance.
However, Ober said, the initial SPC-1 results confirm what many in the industry have long believed, namely that modular storage solutions can offer very good performance at attractive price points when compared with monolithic storage solutions.
The end result is that users still need to determine their specific storage requirements before looking at the SPC's metrics for measuring performance.
"Only once a customer has the answers to these questions can they draw on tools like the SPC to give them meaningful answers," Ober said.
As further SPC-1 benchmark results are announced the differences between tested configurations is expected to narrow, allowing more definitive statements on the performance and price/performance of the respective products to be made, Ober said.
The SPC maintained that this round of benchmarks is the first of many and that future test configurations may include products from the switch vendors.
"That is the nature of the benchmark. It may not be representative of all environments," said Walter Raizner, general manager of storage products for IBM.
Raizner added that there is always an opportunity for other companies to broaden the spectrum of testing configurations.
Let us know what you think about the story, e-mail Kevin Komiega, News Writer
FOR MORE INFORMATION: