Question to the Storage Advisors, from Dark Humour: I just “suffered” through another webinar touting the virtues of SATA without really going into the MTBF or MTBDL (data loss). Does anyone understand TCO at all? I have worked with cheapskate clients and managers who only think about the upfront cost of everything. They don’t plan long term. What I mean is this. The marketeers of this webinar were touting the cost savings of SATA over SCSI in the initial purchase alone.

Usually when someone pinches pennies you pay for it later.

Dear Mr. Humour,

First, I hope the webinar you referred to wasn’t one of mine. I am a big proponent of SAS and it’s ability to support SATA. In fact my last post went into some of the factors that would support SATA having a lower TCO. But I’ve always made it clear in my posts (here, here and here) that the reliability of SATA is much lower than SAS (or SCSI), and therefore one must apply the correct RAID level to overcome that loss in reliability.

In your comment you asked how many SATA drives would it take to equal the reliability of SCSI (or SAS). The question seemed to imply that it would take more SATA drives. But in fact the opposite is true: The more drives you have, the lower the MTTDL of the entire set of drives. So if you want to go from a SCSI RAID-5 array to a SATA array and keep the same or higher reliability, you will either need to (a) lower your total drive count or (b) move to RAID-6. Since SATA drives are so much larger than SCSI drives, it’s possible to decrease the drive count without losing capacity. However since performance is tied to the number of spindles, performance will most certainly drop. I won’t get into RAID-6 since that’s covered in previous posts.

But I do want to address the hot spare comment.

Hot spares are useful for rebuilding the array before a second drive fails. But with today’s huge SATA drives (500GB and climbing), the chance of a bit error during a rebuild is actually more likely than a second failure. In fact, even with drives of equal size, a SATA drive has a 10X probability of a bit error over SCSI.

There are two ways to protect against these bit errors.

The first is to continuously run a background scrub, checking for and correcting bit errors before the array needs to be rebuilt (of course while it’s still optimal). But this degrades performance, and it doesn’t protect against grown defects after a sector is scrubbed.

The second is to use RAID-6, allowing a sector to be rebuilt even in degraded (single drive failure) mode. The downside of RAID-6 is that random write performance will significantly drop.

Thanks for your question. It’s clear that you “get it”.

TT

This entry was posted on Thursday, April 6th, 2006 and is filed under Storage Interconnects & RAID, Advisor - Tom Treadway. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.