Hacker News new | ask | show | jobs
by ck2 4903 days ago
Here's the funny thing about SSD in RAID:

If they are the same model and age, they tend to fail at the same time.

You get more than a single drive failing at the same time = data loss.

(the 100 pushups is darn impressive though)

1 comments

Also true of regular harddrives, just the nature of the game when technically all of the harddrives are under the same load due to RAID and even with different manufacturing times the loads are the same so it's hard to estimate how far apart the batches should be in order to really minimize the incidence of such events.

We do provide snapshots and backups and always recommend that customers backup their servers and take care of ensuring that they have access to their important content in case of any failure.

But maybe we can do the impossible if we can do 100 pushups ;)

The MTBF for SSD in a server environment is two to three years maximum though.

(just ask someone on OVH)

We've seen a high failure rate on SSD compared to SAS & SATA drives. All of our servers are in a RAID environment, so a single failure does not take the system offline. Additionally our backups and snapshots provide another level of redundancy against disk failure. So far no data loss.
That lines up with our lease structure for hardware which is 36mo.

It also lines up with the continuing improvements in server density which compensates for the increased costs in power usage and rack space so it just encourages good behavior in rotating out older hardware.