BackBlaze starts to report on SSD reliability, which they are bigtime.
Click here to post a comment for BackBlaze starts to report on SSD reliability, which they are bigtime. on our message forum
Noisiv
Lets forget for a moment that SSDs are pure electronics, while HDDs are mechanical devices spinning at thousands RPM and held together by PURE MAGIC.
Even without the above bleeding obvious fact and just relying on Backblaze stats I still wouldn't count on SSDs getting much worse down the line.
Why?
Because being only 1 year old on average, SSDs failure rates are negatively affected with the out-of-the-factory phenomena.
Basically some products are stillborns doomed from the very start, and will fail instantly or few months down the line.
Now that they have failed, what we are left with is a healthy population of SSDs. So that that 0.58% AFR for SSDs is not going to get much worse (if any) 12 months from now.
Astyanax
science and engineering*
Noisiv
still needs a shtload of VOODOO 😀
heffeque
Clarke's 3rd law:
Any sufficiently advanced technology is indistinguishable from magic.
nosirrahx
Its good that they are doing this because some people are still convinced that SSDs today have the same reliability as SATA 150 SSDs.
MonstroMart
SSDs are very reliable. I've been running SSDs for as long as the Vertex 2 lineup and so far only my original vertex 2 died. I have an old 120GB Mushkin drive that i will replace soon by a 2TB Samsung one not because it doesn't work anymore but because well 120GB today is nothing lol. I have 6 ssds/nvme in my pc.
schmidtbag
Well, the same is said of HDDs, or really anything electronic. Ever heard of the bathtub curve? In the case of SSDs, the "tub" is just really shallow, and drives that aren't "stillborn" have a reasonably predictable lifespan.
nosirrahx
I have seen 1 NVMe SSD fail ever.
Back when SSDs came to the SATA 150 interface I had a bunch fail.
Truder
Here's a question... Aren't HDDs repairable? As long as the disk isn't damaged, data can potentially be recovered in the event of failure and so drives can be repaired, presenting the option of still being reliable long term storage? (Often HDD failures are due to their controller boards or read/write heads). SSDs on the otherhand have predictable endurance, especially when considering the type of cells used? (QLC have what, 1000 writes per cell?) and over time cells become unreadable (of course the same can be said for hard drives too but that's usually a result of mechanical damage, heads literally damaging a sector for example). Still, I find non-volatile memory fascinating.
While we have here rates of failure over 4 years for HDDs and just 1 year for SSDs, what's to say that after 4 years, SSDs wouldn't fail at increasing rates?
GSDragoon
Not sure how useful this data is without running both types of drives with the same workload.
scoter man1
Isn't the analogy
"if the read/write head were a Boeing 747, and the hard-disk platter were the surface of the Earth:
The head would fly at Mach 800
At less than one centimeter from the ground
And count every blade of grass
Making fewer than 10 unrecoverable counting errors in an area equivalent to all of Ireland."
If that's true, I'd probably break a lot too. In comparison to angry pixies flying around (SSD).
fry178
@Noisiv
it will get worse. not after a year, but probably starts around 2-4y when memory/controller start aging.
at least that was previous data i've seen on ssds.
Ricepudding
I mean it depends on the use, same with anything really... sure if you are writing 1000TB to a 120gb drive a month yeah it won't last long. But I have been using SSD's for the better part of a decade and I have not had one fail ever, and speed wise so its write/read seems consistant even on my 4 years drives I am currently using.
I compare that to my hard drives, and man maybe I as unlucky with them but quite often they would fail in the first few years. Bare in mind I had a lot less useage back then for them and they still would not last as long.
Far as i'm concerned if an SSD doesn't fail in the first month due to something being wrong on the inside it will for the most part last for years and years
Ricepudding
In theory it should increase, but I doubt by much, SSD's least MLC/TLC drives are made to last a long time on normal use. QLC would last less ofc, I do wonder if they should split up the types to show this as just looking at SSD as a whole QLC would have an increase in failture rates. But pretty much SSD's dont tend to fail much since they have no moving parts, its more the aging of the chips inside and how much they can write, considering for an average user they may not reach this for 10-20 years I doubt there will be as much of an increase if we get 4 year data shown
TheDeeGee
Meanwhile a 256GB SSD is dead after 40 days of Chia mining.
Noisiv
well yeah, that's what we're saying
https://en.wikipedia.org/wiki/Bathtub_curve
fry178
@Ricepudding
except your sample size is too small to be relevant.
and i didnt mean write cycles.
it starts around 2y (and up) where controller/chips would die, in numbers 10x of whats expected (~25% vs ~sub 3%).
Orion_13
Dumb question here, how did they calculate the AFR percentage?
Noisiv
Annualized failure rate = probability that the drive will fail in a year = drive failures / drive years
drive years = drive days / 365
