This article can also be found in the Premium Editorial Download "Storage magazine: Storage Products of the Year Awards 2007."
Download it now to read this article plus other related content.
Once upon a time, RAID was good enough for us. But now we deserve better.
RAID was built to improve performance and, more importantly, for protection in the case of drive failures. And because disk drives do three things (read, write and break), RAID has been essential over the years. However, the world has changed radically since RAID was invented. There are storage systems that now minimize the use of RAID and others that are moving away from RAID completely.
RAID was developed at a time when disk capacity was expensive. Ease of management and scalability were secondary. Managing hundreds of terabytes--and even petabytes--is commonplace today; therefore, managing complex RAID environments is becoming untenable. Now that capacity is so cheap and complexity is the bigger problem, we need to shift our perceptions and practices when it comes to RAID. Today, we can afford to create more capacity to protect our data as long as it significantly simplifies storage management.
What's my problem with RAID? It's complex
| and requires too much manual configuration, tuning, analysis and ongoing management. It also creates performance penalties. And if that's not enough, RAID rebuilds can take hours and even days.
If you're wondering which technology should take the place of RAID, my first thought is: Who cares? A storage system should provide availability and uptime with protection against data corruption; users shouldn't have to worry about RAID levels or any other technology that takes its place. That being said, what we do need are storage systems that completely virtualize capacity into logical pools and stripe data across them. Primary data should then be replicated to survive multiple drive failures within the pool. This should all be done transparently and never require a RAID rebuild. As new disk drives are added to the pool, primary and replicated data should be automatically load-balanced to the new capacity as an online and transparent process. This approach is simple and elegant. Some storage folks will think this approach sounds a lot like RAID, but it's the antithesis because RAID today is complicated and messy.
For all those reasons--and from the consumer level (where it's ridiculous to deploy in the first place) to the high end of the market (where sheer magnitude creates the perceived need)--RAID must die. A systems admin from a Fortune 50 firm recently told me he needs an enterprise-class storage system that's as easy to manage as consumer-level products. If that's our goal, we need to change our mindset. Einstein said you can't solve a problem with the same level of consciousness that created it. It may seem crazy to think about eliminating RAID, but we'll never reach necessary levels of ease of use for storage otherwise.
I applaud those next-generation storage system companies that are minimizing or removing the need for RAID. I now call upon the big vendors to follow suit. My advice is simple: You have the power to move the market. You can be responsible for the demise of RAID. Demand high-availability storage systems that don't require RAID, but apply modern technologies built for today's challenges.
This was first published in February 2008