Benchmarking the benchmarks
By Alan Earls
Benchmarks, like EPA mileage ratings, offer an appealing if, at times, flawed means of assessing the relative merit of competing products. Unlike EPA ratings, however, there are numerous different benchmarking standards and tests.
According to Arun Taneja, analyst with Enterprise Storage Group, Milford, Mass., individual switch companies generally use their own benchmarks to measure latency and other characteristics of the switch. The same is true of host bus adapters (HBAs).
However, Taneja says the industry benchmarks that dominate SANs are IOmeter and NetBench while for NAS the dominant benchmark has been SPECsfs from SPEC.org. "EMC and NetApp have fought battles on this for several years --unfortunately, SPEC found a serious bug in the benchmark recently and in mid-June put up a note on their web site that SPECsfs should not be used for NAS comparison until the bug is fixed," says Taneja. This doesn't necessarily mean that SPECsfs numbers are suspect because, Taneja suggests, it may simply have been a problem with accommodating new architectures like BlueArc.
"I think SPEC has the opportunity to take the lead in SANs just like they did in measuring system performance with SPECint and SPECfp," says Taneja. "They already have the organizational and technical infrastructure set up and have relationship with relevant vendor community," he adds.
Additionally, for an alternative view of benchmarks
*Learn more about IOmeter in this searchStorage Administrator tip, "Testing I/O performance".
* We've had some interesting exchange of information on storage benchmarking in our Storage Management Tips and Tricks Discussion Forum. Check it out at http://searchstorage.discussions.techtarget.com/WebX?50@@.ee83ce3.
* And don't forget to send us your own tips! Submit your tip.
About the author: Alan Earls is a freelance writer in Franklin, MA.
This was first published in July 2001