Enterprise -v- Consumer Drives Failure Rates
- Thread starter sboesch
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Interesting read but surely failure rates are not all you pay for when purchasing a more expensive enterprise drive.
Just having a look at the Seagate ES.3 (sata) compared to the Barracuda, read failure rates are listed (if you believe the figures) as an order of magnitue less for the enterprise drive and the seek time is around half.
The fact that a large chunk of their enterprise drives have lasted so well in their data centre environments is still quite surprising. Would add more credability though if the enterprise drives were also run in the same environment so a like for like could be compared or the figures for the number of failures listed which were server based and which were from the pod they put the enterprise drives in and what was the performance differences between enterprise pods and consummer pods.
RB
Just having a look at the Seagate ES.3 (sata) compared to the Barracuda, read failure rates are listed (if you believe the figures) as an order of magnitue less for the enterprise drive and the seek time is around half.
The fact that a large chunk of their enterprise drives have lasted so well in their data centre environments is still quite surprising. Would add more credability though if the enterprise drives were also run in the same environment so a like for like could be compared or the figures for the number of failures listed which were server based and which were from the pod they put the enterprise drives in and what was the performance differences between enterprise pods and consummer pods.
RB
I think they care a lot about performance - in places where it matters. But unlike most they are careful about segmenting their network so that they deploy appropriate technology where it is best suited.
Note that in their other blogs they discuss a segmentation model of storage this way
1. Transactional Storage
2. Bulk Storage
3. Archival Storage
Their use case for consumer drives is focused like a laser onto "bulk storage", where you need high capacity, reasonable performance, high reliability, reasonable resiliency and low cost. Even they use enterprise drives in their own business systems, where transactional performance is more important (yes - I know this blog says its "just because the servers included the drives" - but I don't think that is the whole story).
So assume there are two advantages of enterprise drives that justify their cost premium: performance and reliability. Backblaze does not need their extra performance for "bulk storage". And, at least tentatively, they are debunking the reliability claim. So no need to buy the enterprise drives and pay the premium.
But even they buy the enterprise drives in places where the performance matters.
Note that in their other blogs they discuss a segmentation model of storage this way
1. Transactional Storage
2. Bulk Storage
3. Archival Storage
Their use case for consumer drives is focused like a laser onto "bulk storage", where you need high capacity, reasonable performance, high reliability, reasonable resiliency and low cost. Even they use enterprise drives in their own business systems, where transactional performance is more important (yes - I know this blog says its "just because the servers included the drives" - but I don't think that is the whole story).
So assume there are two advantages of enterprise drives that justify their cost premium: performance and reliability. Backblaze does not need their extra performance for "bulk storage". And, at least tentatively, they are debunking the reliability claim. So no need to buy the enterprise drives and pay the premium.
But even they buy the enterprise drives in places where the performance matters.