I am hearing a lot these days about using cheap technology and transforming it into enterprise reliable storage. I honestly have not done a technology review of any of the multiple vendors and their claims, but I am sure I will at some point soon.
Here are the types of things that I plan on evaluating to better understand the claims.
First, I want to look at the reliability of the underlying disks and connections. Will they be dual-pathed? Will the RAID algorithm rebuild all in a reasonable time? Are the disk AFR and the hard error rates considered? How does the vendor address silent data corruption. Which is more prevalent on lower-end disk storage? How is the whole unit tested compared to enterprise and midrange storage? That would start the discussion.
The second set of questions I would ask would center around failover and failure issues within the controller. Questions such as: How do you access a LUN if part of the controller has failed? How is caching accomplished with and without a failure and most importantly during a failure of something in the hardware path?
The third and last set of questions I will be asking are in the area of performance of the system during normal operations and during various failure scenarios.
As I said, I have not had any exposure to some of the new vendors in this area of yet. There were a number of vendors in the space 5 and 10 years ago that tried doing the same thing, and over time they realized that they had to make the storage controller more robust as the methods and techniques chosen did not work. Of course, everything is different now, but as I have said, lots of things are the same. We shall see.
posted by: Henry Newman