Courtesy: Enterprise Storage Forum
OSD, Declustered RAID Might Help
Some vendors claim that file systems can address the storage reliability problem, but I don’t buy that argument, given the latency inherent in large operating systems and file systems, and the need for high performance (see File System Management Is Headed for Trouble and SSDs, pNFS Will Test RAID Controller Design). And running software RAID-5 or RAID-6 equivalent does not address the underlying issues with the drive. Yes, you could mirror to get out of the disk reliability penalty box, but that does not address the cost issue.
The disk reliability density problem is getting worse, and fast. Some vendors are using techniques such as write logging — keeping track of write on another disk during rebuild to allow the rebuild to occur faster — to get around the growing problem. Will this solve the problem for the long term, or is this the equivalent of the RAID-5 to RAID-6 fix that just delayed the inevitable problem? Personally, I think it will turn out to be just another short-term fix. The real fix must be based on new technology such as OSD, where the disk knows what is stored on it and only has to read and write the objects being managed, not the whole device, or something like declustered RAID. In essence, the disk drive layer needs to have more knowledge of what is storage, or fixed RAID devices must be rethought. Or both.
There are some technologies that are available today and on the way that could help alleviate the problem. The sooner they arrive, the better chance we have of avoiding MTDL.