Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I've always heard that you should not purchase hard disks with sequential serial numbers for any storage system. Now I find myself, for the first time, in a position where I need to purchase around sixteen drives and there seems to be no way to buy this many hard disks and guarantee they weren't all from the same day of manufacturing.

How do I do this? Explain why I'm being overly paranoid.

share|improve this question
    
Great question. –  Wesley Feb 2 '10 at 23:52

4 Answers 4

up vote 12 down vote accepted

I'm going to disagree with joeqwerty here, because I had a similar issue. Check out my experience here, and see the accepted answer. 3/5 of my disks, all purchased in the same transaction, all failed within 24 hours of eachother.

Although I think my experiences were fairly extreme, the real problem happens later on in the life of the device, when all the drives the same age will start to fail in a similar timeframe, so your line of thinking is spot on, but might prove to be more difficult than you think to get around, unless you wait 6 months before buying a 2nd batch of drives.

I guess one solution is to buy disks from different manufacturers and mix them up, however I've had poor experience with this too, because my Western Digitals were 2Mb smaller than my Seagates. Not really an issue when you're building a new array (there will just be 2Mb of unused space on the larger disks), but is a HUGE problem when you've got a failed array and you're trying to rebuild it and you can't becuase all your spare disks are 2Mb too small.

The only thing I can suggest you do is perhaps buy 8 disks from one retailer and then buy another 8 from somewhere else. I doubt you'll be losing out on any volume discounts, and if it helps put your mind at ease then it's worth the extra effort.

share|improve this answer
1  
You can't really argue with experience like that. –  Helvick Feb 2 '10 at 21:10
    
Thanks Dennis. Lose/Loose is my one weakness! –  Mark Henderson Feb 2 '10 at 21:10
2  
Didn't Google do a report stating that failure rates did not correlate to manufacturers or models, but did strongly correlate to production batches? –  Michael Graff Feb 2 '10 at 21:32
3  
This is the paper: labs.google.com/papers/disk_failures.pdf –  touchstone Feb 2 '10 at 22:14
    
I had a similar experience. I populated a server with 16 drives. Within one year, five drives had failed. It has been two years since those five drives were replaced and not a single additional drive has failed. –  David Schwartz Aug 28 '11 at 10:52

Presumably the thinking would be that if there was a glitch in the manufacturing process that affected one drive in a sequentially numbered series (drives from the same manufacturing batch) that all of the drives in that series\batch would be affected. IMHO, there are three reasons to discount this line of thinking:

  1. Modern manufacturing processes are in fact, a science. While manufacturing errors do occur, it seems that the incidence of failure for any x number of units produced is a very small number and that post-manufacturing QA testing would render this number even smaller. The chances of getting some multiple number of units with the same defect seems unlikely to me.

  2. If an anomoly existed in a single unit from a specific batch, that anomoly might slip through testing but I find it hard to believe that a larger number of units with the same inherent anomoly would slip through testing and make it into your hands.

  3. If the issue was an engineering issue and not a manufacturing issue, chances are that all units in the same product family, from any and all manufacturing series\batches would be affected so it wouldn't really matter if they came from the same batch or not, they would all be faulty until the engineering problem was discovered, resolved, and the manufacturing process adjusted accordingly.

Of course, I'm neither a manufacturing nor an engineering expert, so these are just my thoughts on the matter.

share|improve this answer
    
Yet, despite all that, manufacturers do indeed produce bad batches of drives. –  John Gardeniers Feb 3 '10 at 3:15

joeqwerty's answer is great on the 'why', which I think you already understand. As for the 'how', you ask your retailer or reseller, and if they can't help you, you spend a little bit more and go to someone who can.

Alternatively, you can buy different brands, which is a good way to assure that you won't at least have all your disks fail at once.

share|improve this answer

You are not paranoid, it's indeed very bad to use drives from the same batches, for a simple reason :

  • Same drives, built the same way, from the same components, used the same way, are likely to fail at a same time.

I've been working for years at a data recovery company, and i've seen amounts of RAID array with several drives failing at the same time (within array rebuild time), usually all those with consecutive or very close serial numbers (I recall a sad customer who had a RAID of 5 drives, 3 of them having the magnetic substrate wiped out from the platter...)

I'm not sure if you are talking about a RAID system here, but if so, I advice against mixing drives from different makers, as you'll lose performance due to different seek times and data transfer rate, and as Farseeker stated, you're likely to have size issues with raid.

To splay the failure date of these drives, you should turn to different resellers, since it's unlikely they'll have the same batch of drive.

You can also "burn" those coming from the same batch : use heavily one drive alone one week, add another drive and use them both heavily another week, add a third drive and use them three heavily ... For 16 drives it is a bit inconvenient, so you should evaluate the risks of the drives failing in the same period to set the burning time.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.