I have been a bit neglectful of writing about the 32 types of storage ownership costs. This entry is an extension on item #26 - Storage Performance as a cost of ownership.

Many may see performance cost as a soft cost, and it can be hard to provide cost savings or expenses unless we look at comparative economics. In comparative economics, one solution may be must faster (IOPS) than a second solution, so the opportunity cost or opportunity loss can be factored into the TCO comparison.

Performance and scalability have a tangential but measurable impact on expenses. Storage performance ties directly to end-user productivity and may help to contribute to deferring future hardware purchases: if more storage is available at the appropriate levels of performance needed by applications, storage managers need not buy further high-end storage resources to satisfy capacity growth.

Efficiently matching business applications performance demands and storage infrastructure capabilities is a key tenet of  tiered storage infrastructures. The scalability of a storage system affects cost in the same way that performance does. If the storage systems invested in do not hold to their performance characteristics as capacity increases, it will be necessary to purchase more high-end capacity to satisfy the needs of applications.

 A recent case-in-point, a client/friend of mine works for a major airlines, and when architecting a new server/storage infrastructure for the web-based reservation system, the total IOPS rating was used and dollarized to show the revenue potential difference between competing systems.

I talk with many people who feel that $/IOPS is a more meaningful metric than $/GB in making storage decisions, and economic comparisons for storage. Performance tends to me a more critical measure of system design, and if cheaper disk (high density SATA) are used, more spindles may be needed to sustain the throughput. It will take more SATA drives to maintain a certain performance level, irrespective of the disk spindle capacity. That is why $/IOPS or IOPS/$ is a better metric in storage planning.

Trackback URI | Comments RSS