3 ms·

Impressed with the size, I need to see what size IBM is offering us for our next move, but we go from 6 racks of 15k drives to one rack. While the SSD cost in

Impressed with the size, I need to see what size IBM is offering us for our next move, but we go from 6 racks of 15k drives to one rack. While the SSD cost in our setup is more expensive than spinning media IBM is pushing it and the price has dropped immensely since first availability. The second bonus we have is we have systems at COLOs which charge both for power, cooling, and rack space. Needless to say those all come down and since our processing heavily batch oriented with strong web based activity the SSD systems will certainly boost performance and provide a good two to three year buffer. Fortunately Toshiba and Seagate are moving the density so fast that by next upgrade the prices should be amazingly different (we are 60tb, but mirrored meaning 120tb). Now I am curious if our AIX side will have fun here as they should make for easier flash copies that would be faster too. I guess my real concern is, that is a lot of data to lose to one device failure. So raid/mirroring is so much more important.

atYevP 10y ago
> I guess my real concern is, that is a lot of data to lose to one device failure. So raid/mirroring is so much more important. Yev from Backblaze here -> Yea, we raid across different pods w/ Reed Solomon (https://www.backblaze.com/blog/reed-solomon/ https://www.backblaze.com/blog/reed-solomon/). With these that would be very costly, but the benefit for us is that an entire pod or two can go down and it's still OK.
- mixedCase 10y ago
  How would you say your solution compares to groups of 10 drives in RAID6 configuration when it comes to safety, storage and availability?
  - atYevP 10y ago
    You can read a bit more about our infrastructure -> https://www.backblaze.com/blog/vault-cloud-storage-architecture/ https://www.backblaze.com/blog/vault-cloud-storage-architect... - Right now we essentially "RAID" across multiple racks - so we can lose an entire rack and it wouldn't really affect us too much. Having 10 drives in RAID6 is great - having those 10 drives be in different physical locations on separate power circuits is even better :D

atYevP 10y ago

> I guess my real concern is, that is a lot of data to lose to one device failure. So raid/mirroring is so much more important. Yev from Backblaze here -> Yea, we raid across different pods w/ Reed Solomon (https://www.backblaze.com/blog/reed-solomon/ https://www.backblaze.com/blog/reed-solomon/). With these that would be very costly, but the benefit for us is that an entire pod or two can go down and it's still OK.

mixedCase 10y ago
How would you say your solution compares to groups of 10 drives in RAID6 configuration when it comes to safety, storage and availability?
- atYevP 10y ago
  You can read a bit more about our infrastructure -> https://www.backblaze.com/blog/vault-cloud-storage-architecture/ https://www.backblaze.com/blog/vault-cloud-storage-architect... - Right now we essentially "RAID" across multiple racks - so we can lose an entire rack and it wouldn't really affect us too much. Having 10 drives in RAID6 is great - having those 10 drives be in different physical locations on separate power circuits is even better :D