Linked by Thom Holwerda on Mon 18th Jan 2010 16:57 UTC, submitted by wanker90210
Hardware, Embedded Systems ACM's latest journal had an interesting article about RAID which suggested it might be time for triple parity raid. "How much longer will current RAID techniques persevere? The RAID levels were codified in the late 1980s; double-parity RAID, known as RAID-6, is the current standard for high-availability, space-efficient storage. The incredible growth of hard-drive capacities, however, could impose serious limitations on the reliability even of RAID-6 systems. Recent trends in hard drives show that triple-parity RAID must soon become pervasive."
Permalink for comment 404747
To read all comments associated with this story, please click here.
ACM triple parity article
by cjcox on Mon 18th Jan 2010 21:45 UTC
cjcox
Member since:
2006-12-21

The gist of the article was that a RAID7 should be defined for 3 or MORE parity.

But in all fairness, I believe the answer is rethinking storage strictly in terms of mirrors. So I believe that object based storage is the future.

The idea being that you have something called "storage" and it's constructed out of mirrored "elements". Mirrored elements are ideally created using the best independent paths possible. As failures occur, elements that are affected are re-mirrored, again to the best independent path. Such a system usually requires independently redundant meta data areas for fast element lookups, etc.

A highly reliable object storage system would use 3 element mirrors (for example).

Sure, there's more space used to provide extra reliability, but such a system will also likely greatly outperform write parity approaches on both read (if done well) and write.

Also, these kind of object based storage systems should be able to scale well beyond petabytes ideally (something that is hard to do today).

Rebuilds on parity based RAID can take days... and it's getting worse. With good granularity, object based systems might rebuild in a matter of seconds or just a few minutes.

Reply Score: 1