So I had a random thought about what is the most platters that any hard drive has had. I looked it up it seems that the Western Digital Ultrastar® DC HC690 has eleven platters in a 3.5” form factor. That certainly gives you a lot more bandwidth, though not much help for seek time (unless you do the half-allocated trick).

I have a 13 years old NAS with 4x1TB consumer drives with over 10y head flying hours and 600,000 head unloads. Only 1 drive failed at around 7 years. The remaining 3 are still spinning and pass the long self test. I do manually set the hdparm -B and -S to balance head flying vs unloads, and I keep the NAS in my basement so everything is thermally cool. I'm kinda of hoping the other drives will fail so I can get a new NAS but no such luck yet :-(

I'm curious what this data would look like collated by drive birth date rather than (or in 3D addition to) age. I wouldn't use that as the "primary" way to look at things, but it could pop some interesting bits. Maybe one of the manufacturers had a shipload of subpar grease? Slightly shittier magnets? Poor quality silicon? There's all kinds of things that could cause a few months of hard drive manufacture to be slightly less reliable…

(Also: "Accumulated power on time, hours:minutes 37451*:12, Manufactured in week 27 of year 2014" — I might want to replace these :D — * pretty sure that overflowed at 16 bit, they were powered on almost continuously & adding 65536 makes it 11.7 years.)

When I am projecting prices I tend to assume a 5 year life for a consumer hard drive. I do wonder from this data and the change in purchasing from backblaze if the enterprise class drives might pay for their extra price if they survive out to more like 9 years. 20% extra cost per TB verses about 30%+ more life time. They do tend to consume a bit more power and make more noise as well. I wish they had more data on why the drives were surviving longer, if its purchasing in palettes there isn't a lot we can do, but if its that enterprise drives are a lot better than NAS or basic consumer drives then that we compare cost wise.

Personal anecdote - I would say (a cautious) yes. Bought 3 WD hard drives (1 external, 2 internal, during different time periods; in the last 10+ years) for personal use and 2 failed exactly after the 5 year warranty period ended (within a month or so). One failed just a few weeks before the warranty period, and so WD had to replace it (and I got a replacement HDD that I could use for another 5 years). That's good engineering! (I also have an old 500GB external Seagate drive that has now lasted 10+ years, and still works perfectly - probably an outlier).

That said, one thing that I do find very attractive in Seagate HDDs now is that they are also offering free data recovery within the warranty period, with some models. Anybody who has lost data (i.e. idiots like me who didn't care about backups) and had to use such services knows how expensive they can be.

Per charts in TFA, it looks like some disks are failing less overall, and failing after a longer period of time.

I'm still not sure how to confidently store decent amounts of (personal) data for over 5 years without

  1- giving to cloud,
  2- burning to M-disk, or
  3- replacing multiple HDD every 5 years on average

Who has the easy, self-service, cost-effective solution for basic, durable file storage? Synology? TrueNAS? Debian? UGreen?

(1) and (2) both have their annoyances, so (3) seems "best" still, but seems "too complex" for most? I'd consider myself pretty technical, and I'd say (3) presents real challenges if I don't want it to become a somewhat significant hobby.

> The issue isn’t that the bathtub curve is wrong—it’s that it’s incomplete.

Well, yeah. The bathtub curve is a simplified model that is ‘wrong’, but it is also a very useful concept regarding time to failure (with some pretty big and obvious caveats) that you can broadly apply to many manufactured things.

Just like Newtonian physics breaks down when you get closer to the speed of light, the bathtub curve breaks down when you introduce firmware into the mix or create dependencies between units so they can fail together.

I know the article mentions these things, and I hate to be pedantic, but the bathtub curve is still a useful construct and is alive and well. Just use it properly.

I feel like I’d like to see graphs in the shape you see in some medical trials – time on the x axis and % still alive on the y. You could group drives by the year they were purchased and have multiple lines for different years on there.

Ah I haven’t seen the yearly backblaze post in some time now, glad it’s back.

Hard drives are not getting better.

Hard drives you can conveniently buy as a consumer - yes. There's a difference.

Not from the prices I'm seeing.

Do we have enough rare earth metals to provide storage for the AI boom?

pleasant contradiction to betteridge's law