At this juncture in the time the emerging technology that is capturing the most amount of attention from CIOs these days is storage virtualization.
That should not come as too much of a surprise given because storage virtualization seems to naturally follow from server virtualization.
But storage virtualization is a whole lot harder than server virtualization because there are a lot subtle issues concerning performance that IT folks have to take into consideration. Foremost among those issues is performance. If you start to embrace virtual storage you are essentially combining different data sets on to the same storage array. But one of the guiding principles of application performance that has developed over the years is that each application set performs a lot better when it has its own dedicated storage array.
This situation creates a conundrum for IT people because on the one hand they want to reduce the number of physical arrays they need to manage and pay for, but on the other hand their loathe to do anything that might have a negative impact on the I/O architecture that drives application performance.
Interestingly enough, the answer to this riddle might have nothing to do with virtualization and everything to do with compression. Just about every storage vendor out there has some sort of compression algorithm that it has implemented but the question is how efficient are these algorithms when they are written by companies that make their money by selling storage arrays in volume.
That's the challenge that a new storage startup company called Ocarina Networks has taken on. The company has developed an appliance that first decompresses all the files that have been compressed using a storage vendor's algorithm and then compresses them again using an algorithm developed by Ocarina to store those files more efficiently on the existing storage arrays. As part of the effort, the Ocarina software can also detect similarities in, for example PowerPoint presentations, so it needs to save fewer versions of the various PowerPoint presentations because in actual reality the different versions have only minor differences that Ocarina can detect.
This means that the customer can increase the utilization by up to a factor of 10 on any given storage array without necessarily having to resort to virtualization to accomplish that. At the moment, the Ocarina Networks appliance costs about $75,000 and adds about 4 milliseconds of access time overhead to the storage performance equation. But because the files are now stored in a file format developed by Ocarina Networks a special reader that Ocarina intends to make available for free is required.
Historically, storage vendors have not had much of an incentive to invest in developing new ways to store files more efficiently on existing arrays because they are in the business of selling more arrays. Even though virtualization may ultimately leads to few unit sales vendors can still get behind it because you need to buy a new array to take advantage of it.
This is why it will be interesting to watch how storage vendors will react to Ocarina Networks once the company starts shipping its appliance in the fall. In the meantime, IT organizations need to figure out if they are simply trying to increase utilization rates of their existing storage or whether they need a whole new approach to data management that might truly require storage virtualization.