RAID5 May Soon Be Obsolete
Posted 10/25/08 at 03:35:07 PM by Justin Kerr
RAID 5 users anxiously awaiting the debut of 2 TB drives to help build massive storage array’s may want to think twice before taking the plunge. An in-depth look into the underlying problems with massive storage RAID5 configurations suggests that s a single drive as redundancy might not cut it anymore. SATA drives carry a specified unrecoverable read rate of 10^14. This might sound like a huge number, but it basically tells us that any array in excess of 11.37 TB will contain at least one unrecoverable read. In the case of a RAID 5 rebuild, this can be catastrophic.
Storage companies want us to believe that RAID 6 will address the issue but sadly this doesn’t seem to be the case long term. The additional drive will increase redundancy, but since failures will always be followed by read errors on another drive, RAID 6 won’t help you one bit (pardon the pun). Many of the advances in storage capacity are a result of perpendicular recording which help reduce this problem, but it still exists. Perpendicular recording also isn’t likely to take us much further beyond 2TB. Delays in finding new methods will force those in need of massive storage configurations to turn to RAID5 or 6 which as we know now, are vulnerable.
Don’t get me wrong, we still have time to come up with a solution to this problem, but users stringing together 6 or more disks as early as next year could start hitting this ceiling. Additionally, in order for the worst case scenario to occur (loss of all data) the one failed drive would need to be completely unrecoverable. So the sky isn’t falling, but it sure makes for an interesting problem. Want to learn more about RAID? Read Maximum PC’s in-depth guide, RAID done Right.
An interesting item of note, on human progress. The picture shown below is just slightly over 2TB of storage. This much capacity will be possible in a single drive by next year. Isn’t progress wonderful?

(Image Credit Rainwulf.com)
RAID-Z?
Submitted by icmb on Sat, 10/25/2008 - 3:09pm
What about the RAID-Z in Sun's ZFS filesystem on Solaris? Doesn't that avoid this problem?
i was wondering if you
Submitted by dethdeks on Sat, 10/25/2008 - 3:03pm
i was wondering if you noticed this or not but you were talking about sata drives but yet the picture of hdd's were all Pata drives, just wondering if this was done on purpose or just by accident
SCSI not PATA
Submitted by praetor_alpha on Sun, 10/26/2008 - 9:22am
Those are probably all SCSI drives, taken straight out of a server somewhere. PATA drives probably wouldn't be able to scale that far, right?
They're all PATA there. If
Submitted by icmb on Sun, 10/26/2008 - 10:43am
They're all PATA there. If you follow the link it says exactly what they are. The purpose of the pic was just to show what 2 TB *used* to take. It's unrelated to the SATA/RAID issue itself
You are correct
Submitted by Justin.Kerr on Sun, 10/26/2008 - 10:59am
Couldn't have said it better myself.
Feature
Review
Feature
Feature
Feature






