I have ranted (a regular occurrence) for years over the fact that standard RAID will not last forever. I last seriously ranted over two years ago. The article was widely read and widely commented on, as I got many emails. Two years later and most agree with me, but where are the changes? We now have 3 TB drives with 4 TB drives on the horizon.
The timebomb of RAID and data loss is nearing. I have heard of a number of cases where multiples failures happened on large file systems during rebuild. When 4 TB drives come with the typical 20 percent increase in performance and now 33 percent (3TB to 4 TB drives) increase in density, the problem is going to happen more and more often. Declustered RAID is the solution to the problem, but it is slow in appearing in the market.
Some vendors that have addressed the problem. Since I try not to mention vendor names, I will let you ask the questions of the community vendors yourselves. I think market pressure is the only way to get the vendor community to make these kinds of major technology changes. All of the vendors I have talked to over the years understand and agree with my conjecture that things must change. However, change is expensive, and we had a recession in which vendors did not invest in technology for that very good reason. The economic situation is seemingly getting better, and it is time for the user community to demand the vendor community meet our reliability requirements. With 4 TB drives just around the corner, and hard error rates still at the same 1 sector in 10E15 bits for the more than six years, the time for change is now.
Labels: RAID,data loss prevention,data loss
posted by: Henry Newman