It's World Backup Day! Rsync + SSH + Synology For Easy Secure Backup | Level One Techs

Learn more about data backup: http://sy.to/k4aho

Enter to win a Synology NAS, HDDs and more: http://sy.to/f8fwn


This is a companion discussion topic for the original entry at https://level1techs.com/video/its-world-backup-day-rsync-ssh-synology-easy-secure-backup
2 Likes

I get mechanical hdd failures (moving parts and friction are not your friend), but how often do ssds fail?

1 Like

All drives die.
But a quality SSD will generally last longer than it’s warranty/TBW number.

It’s not the nand cells that are usually the issue; a good controller will share the load among the chips, with wear levelling. It’s the other parts like the controller or whatever that seems to fail.

1 Like

Yeah I know all ssds have a limited amount of data cycles they can read and write. But what usually kills the drives? Lack of cooling? Like what usually goes wrong with a NAS SSD?

1 Like

That is a good question.

It depends on the drive.

On some, the controller goes bad.

On some, the NAND goes bad.

With most SSDs though, you aren’t going to overheat them as long as they’ve got a bit of airflow (even hot air from a GPU)

some enterprise ssds overflow if they have been on for 32,768 hours and then brick. :smiley:

2 Likes

That sounds like a feature, not a bug.

2 Likes

Didn’t some enterprise SSD’s lack TRIM, and kinda fill up, then drop performance?
Fine after a wipe, but needed a little TLC to get back up to speed?

Lot of raid controllers STILL don’t support trim…

3 Likes

Which is why jbod is always best.

trim can expose information about encrypted drives on Linux (don’t know about Windows) Fedora 32 will come with a trim enabled systemd unit…

They used to be pretty bad because of heavy writes. Now they have wear leveling and usually come in odd sizes so that the controller can assign bits to parts of the drive with less wear to maintain drive health.

I honestly think drives deployed for the OS can last a VERY long time (maybe a decade) Since they will just read a ton of data. Writes are what wear a drive out. A teat a few years ago by Tech Report wrote the Full Capacity of several drives until the drives died.

The majority of the drives lasted 18 months writing roughly 240GB every day, the drive that won, wrote 2.5 Petabytes until it became unwriteable, but could be read. This was 4 years ago btw.

note: That is Scott Wason before he joined AMD/RTG