5 ms·

I read the article with great anticipation, but the author fails to poke any more holes in Backblaze's test that they themselves haven't already caveated. Back

by pcurve 13y ago

I read the article with great anticipation, but the author fails to poke any more holes in Backblaze's test that they themselves haven't already caveated. Backblaze already acknowledged that their usage environment doesn't reflect that of normal consumer, yet the author harps on the point. The author tries to debunk Backblaze's claim that it found no correlation between drive temperature and failure rate. Yet he presents no evidence that Backblaze's methodolgy was wrong. Backblaze had large enough sample to have even distribution of drives in various temperature, rack, and load environment. I doubt they put all Seagate drives in the middle shelf of Ver. 1 racks where it was the hottest, but put all Hitachi drives in the bottom of the Ver. 3 racks where it was cooler and more stable. Unless there is another study that shows failure rate of these drives in normal consumer environment, Seagate's 3-5x failure rate should give people a pause.

wiredfool 13y ago
I find it particularly galling that "Everybody knows temperature is important" is the rebuttal to a bunch of data saying "we didn't see an effect". If you've got data, post it. If not, it starts to sound like astroturf. Fwiw, with my small sample size, I agree with backblaze. Seagates appear to be failing more, and avoid wdc greens.
- zzzcpan 13y ago
  > and avoid wdc greens Why? Backblaze reports 3.6% AFR on wdc greens over 4.4 years, which is pretty good.
  - wiredfool 13y ago
    I think they had issues with early death from vibration, and they don't perform well in anything like demanding performance situations. Though that may just be my faulty memory of he The original piece.
  - sexmonad 13y ago
    Anecdotally, at my last job, we had used them on several of our internal file and VM servers (because they were cheap), and they had a nasty habit of falling off the bus overnight, causing the RAID controller to go berserk.
mindslight 13y ago
I love how he keeps coming back to how the drives were sourced through retail channels - wtf does he think consumers get them from? I frankly don't care if one brand lasts longer when handled with white gloves, because I'm not going to be paying someone to pretend to do that for me. Also, insisting that Backblaze has some ridiculously out-of-this-world "workload"? We're not talking a database, but an operation that is usually done with tapes. According to him, consumers must only need 4TB drives with 1000 hour MTBFs and 20TB of lifetime IOs.
sliken 13y ago
I thought the point about which version chassis the drives in to be a valid point. I believe the backblaze design has 1 pound drive vibrating at 120hz or so supported by a SATA connector. I don't know the SATA specification, but I suspect they aren't designed to support the drives weight. If SATA connectors where designed to support the drive they wouldn't be plastic. Additionally the SATA Multiplier boards (thin fiberglass with delicate traces) are supporting the weight of 5 drives each. This would also explain why the enterprise drives (typically heavier, more platters, and with a more conservative design) are less reliable then the consumer (with typically higher density platters).

by pcurve 13y ago

wiredfool 13y ago

I find it particularly galling that "Everybody knows temperature is important" is the rebuttal to a bunch of data saying "we didn't see an effect". If you've got data, post it. If not, it starts to sound like astroturf. Fwiw, with my small sample size, I agree with backblaze. Seagates appear to be failing more, and avoid wdc greens.

zzzcpan 13y ago
> and avoid wdc greens Why? Backblaze reports 3.6% AFR on wdc greens over 4.4 years, which is pretty good.
- wiredfool 13y ago
  I think they had issues with early death from vibration, and they don't perform well in anything like demanding performance situations. Though that may just be my faulty memory of he The original piece.
- sexmonad 13y ago
  Anecdotally, at my last job, we had used them on several of our internal file and VM servers (because they were cheap), and they had a nasty habit of falling off the bus overnight, causing the RAID controller to go berserk.

mindslight 13y ago

I love how he keeps coming back to how the drives were sourced through retail channels - wtf does he think consumers get them from? I frankly don't care if one brand lasts longer when handled with white gloves, because I'm not going to be paying someone to pretend to do that for me. Also, insisting that Backblaze has some ridiculously out-of-this-world "workload"? We're not talking a database, but an operation that is usually done with tapes. According to him, consumers must only need 4TB drives with 1000 hour MTBFs and 20TB of lifetime IOs.

sliken 13y ago

I thought the point about which version chassis the drives in to be a valid point. I believe the backblaze design has 1 pound drive vibrating at 120hz or so supported by a SATA connector. I don't know the SATA specification, but I suspect they aren't designed to support the drives weight. If SATA connectors where designed to support the drive they wouldn't be plastic. Additionally the SATA Multiplier boards (thin fiberglass with delicate traces) are supporting the weight of 5 drives each. This would also explain why the enterprise drives (typically heavier, more platters, and with a more conservative design) are less reliable then the consumer (with typically higher density platters).