As a Network Engineer for Collages.Net I handle storage - lots of it. While the figures of yesterday, even on the enterprise level was oftentimes gigabytes or a few terabytes, now dozens or hundreds of terabytes is common-place, even amongst smaller businesses. This has raised new challenges and opportunities for the storage industry. Robin Harris recently wrote an article over at ZDNet entitled Why RAID 5 stops working in 2009. While the title is a bit misleading, the article itself is dead on. RAID 5 is no longer a workable solution for many businesses. For those who don't understand the concept of RAID - let me explain. RAID creates redundancy across storage. It ensures that if a single drive fails the data is still available. One popular method of doing this is RAID 5, which uses a parity method. This means that across all the disks in a RAID 5 array there is parity data. This parity data allows a drive to fail and yet for the data to still be restored. This is accomplished using mathematics. a + b = c. You can solve for either a, b, or c such as a + ? = c, c-a = ?. However, when two drives fail, you are out of luck.

    The problem, as Harris points out, is that because of the increasing size of disks it is more likely that you will experience a total failure rather than just a single drive failure. This is created by a number of factors we don't need to look at. What we do need to discuss is the solution. One popular solution is RAID 6. This allows for two drives to be lost before there is any data loss. Good, but as Harris aptly points out, drives will continue to expand in size and at some point RAID 6 will no longer be an equitable solution.

    Google (and several other large enterprises) solution to this issue has been to abandon RAID, or, perhaps more correctly - to spread RAID across multiple systems. Thus a unit itself doesn't have a RAID, rather the data is kept on multiple servers - thus even if several fail there are still several remaining.

    Unfortunately, these sort of solutions are not available to most corporations without significant in-house development, or the utilization of hyper-expensive third party solutions. There is great room within the industry for innovation in this arena...but I'm tired of writing this now, so I digress.