Storage Adviser | Mario Apicella » TAG: Open discussion

May 16, 2007 | Comments: (0)

More expert opinions on disk drives reliability

TAGS: Open discussion

There's a paper on getting RAID reliability with up to 3% HDD failure rates.

comments Gordon Hughes, perhaps in response to this Storage Insider column.

I wrote it based on my experience that field failure rates are this high, after the infant reliability demonstrated by the drive makers. (0.7% AFR is based on just a few months of testing a few thousand drives of a new model. This testing removes the infant mortality part of the FAR bathtub curve, as the CMU FAST paper shows.) Years ago, Net App showed 3% AFR data on their SCSI drive arrays.

For an explananion of those acronyms please see Gordon's paper.

My paper is: "Reliability and Security of RAID Storage Systems and D2D Archives using SATA Disk Drives" ACM Transactions on Storage, December 2004.

Very interesting reading, while sipping a cup of java or two. At the end of that paper Gordon suggests further studies from which:

A class of “Enterprise SATA” drives might emerge, still allowing the high SATA drive capacities at acceptable cost

which is exactly what happened.

Happy reading!

Posted by Mario Apicella on May 16, 2007 07:41 AM

October 25, 2006 | Comments: (0)

Is EMC focusing too much on software?

TAGS: Open discussion

I have mixed feelings about the latest flow of announcements from EMC. EMC makes some interesting promises on the hardware front, including a greener Symmetrix box and Clariion arrays that can walk and chew gum, I mean handle iSCSI and FC traffic at the same time.

What I find suprising is that EMC has not yet implemented RAID 6, a feature already embraced by many vendors that deploy SATA drives. You may remember a discussion we had in June on this blog about SATA and RAID 5 .

A quick wrap up of that discussion is that RAID 6 can reduce the vulnerability caused by the failure of a second drive when one is akready being re-built.

Obviously, larger drives trigger a longer rebuild time hence data loss if a second drive breaks during that time and you have RAID 5. RAID 6 controllers implement dual parity, so can survive the simultaneous failure of two drives.

In all fairness, I haven't tested in person how long does it take to rebuild a SATA drive on a Clariion but other vendors have measured on their own hardware a 4 to 12 hours elapsed time or longer.

Is EMC focusing too much on software and acquisitions and loosing touch with hardware?

Posted by Mario Apicella on October 25, 2006 06:55 AM