Your answer to the Storage system benchmarks question (26 July 2001) provided a great list of the many basic factors that require consideration. I especially liked your comment that "The best way to determine if a specific vendor?s solution is right for you is to do an evaluation using your real world production environment and test it."
Along these lines, I would think that the ability to quickly and easily determine the actual performance of your files (at the file level, with empirical I/O performance metrics) would be a great help. I'd be interested to know if you agree.
Incidentally, you might want to find out about the hIOmon File I/O Performance Monitor from hyperI/O LLC ( http://www.hyperio.com/).
Real world I/O testing that validates production environments is always a better test than running benchmarks. When doing that testing, if actual file level I/O statistics can be gathered, then all the better.
RAW throughput and RAW IOPS statistics are great to get a general idea of how your production applications should perform on a given platform. Standardized benchmark tests can give you this information, and are a great tool in TEST environments when considering platforms from different vendors. If production applications can be used in place of the benchmark tests, then your statistical data can be used to compare actual current production environments, which is always a better indicator of value.
Thanks for the tip on hyperIO.
This was first published in August 2001