MTBF

[StefanN]

Are there any figures for "Mean Time Between Failures" not for disks but for nodes ore whole clusters?

[hariprak · Accepted solution]

Hi,

Ideally, it should be 99.999%, Please remember that we deploy clusters to avoid downtime,not failures. MTBF is for individual hard drives.

Thanks

[NAYABSK]

Hi Stefan, 

The most commonly referenced MTBF values for storage subsystem devices are for drives. SSD, SATA,

SAS, and FC drives have different MTBF values, as follows:

--> SSD (SLC) drives are 2.0 million hours

--> SAS and FC drives are 1.6 million hours

--> SATA drives are 1.2 million hours

The drive warranty is five years (43,800 hours), which is far short of 1.6 million or even 1.2 million hours.
Again, MTBF is a measure not of the usable life of the drive but rather of the error rate within the useful drive
life.

Purely based on MTBF, the math suggests that for SATA drives (1.2 million hours MTBF) ~0.73% of your
deployed drives should fail each year. For FC and SAS drives (1.6 million hours MTBF) ~0.55% of your
deployed drives should fail each year. For SSD drives (2.0 million hours MTBF) ~.44% of your deployed
drives should fail each year.

Thanks,

Nayab

**** If my solution helped you, Please mark it as solution to help others****