Hacker News Reader

Comment by mvanbaak | Hacker News Reader

the 'secret' is not that you turn them off. it's simply luck.

I have 4TB HGST drives running 24/7 for over a decade. ok, not 24 but 8, and also 0 failures. But I'm also lucky, like you. Some of the people I know have several RMAs with the same drives so there's that.

My main question is: What is it that takes 71TB but can be turned off most of the time? Is this the server you store backups?

View on HN

In fact, the conventional wisdom for a long time was to not turn them off if you want longevity. Bearings seize when cold for instance.

It can be luck, but with 24 drives, it feels very lucky. Somebody with proper statistics knowledge can probably calculate the risk with a guestimated 1% yearly failure rate how likely it would be to have all 24 drives remaining.

And remember, my previous NAS with 20 drives also didn't have any failures. So N=44, how lucky must I be?

It's for residential usage, and if I need some data, I often just copy it over 10Gbit to a system that uses much less power and this NAS is then turned off again.

We don't really have to guess. Backblaze posted their stats for 4 TB HGST drives for 2024, and of their 10,000 drives, 5 failed. If OP's 2014 4 TB HGST drives are anything like this, then this is just snake oil and magic rituals and it doesn't really matter what you do.

> If OP's 2014 4 TB HGST drives are anything like this, then this is just snake oil and magic rituals and it doesn't really matter what you do.

It might matter what you do, but we only have public data for people in datacenters. Not a whole lot of people with 10,000 drives are going to have them mostly turned off, and none of them shared their data.

5 drives failed in Q1 2024. 8 died in Q2. That is still a very low failure rate.

Drives have a bathtub curve, but if you want you can be conservative and estimate first year failure rates throughout. So that's p=5/10000 for drive failure. So chance of no-failure per year (because of our assumption) is 1-p. So, chance of no-failure per ten year is (1-p)^10 or about 99.5%

Is that for one drive, or for all 24 drives to survive 10 years?

That's for one drive. For all 24 it's about 88.7%.

Those are different drives though, they're MegaScale DC 4000 while OP is using 4 TB Deskstars. Not sure if they're basically the same (probably). I've also had a bunch of these 4TB Megascale drives and absolutely no problems whatsoever in about 10 years as well. Run very cool as well (I think they're 5400 rpm not 7200 rpm).

The main issue with drives like these is that 4 TB is just so little storage compared to 16-20 TB class drives, it kinda gets hard to justify the backplane slot.

it's (1-p)^24^10, where p is drive failure rate per year (assuming it doesn't go up over time). so at 1% that's about 9% or a 1/10 chance of this result. Not exactly great, but not impossible.

the backblaze rates are all over the place, but it does appear they have drives that this rate or lower: https://www.backblaze.com/blog/backblaze-drive-stats-for-q2-...

I can see that, my assumption that my result (no drive failure over 10 year) was rare is wrong. So I've updated my blog about that.

The failure rate is not truly random with a nice normal distribution of failures over time. There are sometimes higher rates in specific batches or they can start failing altogether etc.

Backblaze reports always are interesting insights into how consumers drives behave under constant load.

I'm curious what the "Stop/Stop cycle count" is on these drives and roughly how many times per week/day you are accessing the server.

> What is it that takes 71TB but can be turned off most of the time?

Still waiting for somebody to explain this to me as well.

I have a 22TB RAID10 system out in my detached garage that works as an "off-site" backup server for all my other systems. It stays off most of the time. It's on when I'm backing up data to it, or if it's running backups to LTO tape. Or it's on when I'm out in the garage doing whatever project, I use it to play music and look up stuff on the web. Otherwise it's off, most of the time.

Reply on news.ycombinator.com