End of RAID-5?
Robin Harris of StorageMojo verbalized an interesting theory regarding an imminent inpractiability of RAID5 within 2 years:
The arithmetic goes like this. Take a 7 drive RAID 5 stripe. Each drive is 2 TB (in a couple of years). One drive fails, leaving 12 TB of capacity to read to recreate the lost data. With a SATA URE of 10^14, which is about 12 TB - OK, a little more - you are highly likely to encounter a URE. At that point an honest RAID controller will inform you that it can’t complete the rebuild.
My stance to this topic is a little bit different. In my personal opinion, Filesystems and RAID technology without strong checksum will be unacceptable. You doesn´t need RAID6 because of the URE (unrecoverable error) problem, when you have different means to ensure data integrity. By having checksums, you can do stuff like combinational reconstruction of a RAID set (from the ZFS FAQ):
RAID-Z utilizes the ZFS checksum mechanisms to prove the integrity of the data before handing it back to the applications. In the event of detecting corrupted data, the RAID-Z logic can combine checksums with the parity information to to not only return the correct data, but also to determine the corrupted data and correct it. This combinatorial reconstruction is impossible with RAID-5. It is a feature unique to ZFS stemming from the fact that both traditional filesystem and volume manager functionality has been integrated into a more intelligent whole.
Thus RAID-Z can give Sun an advantage: When ZFS can deliver RAID-6 availability with a RAID-5 equivalent amount of disk, this saves spindles, thus power and money.