the issue is not max capacity. the issue is TB per dollar. HDD have been quite stagnant since 2010 and will do even worse in 2021 with the miniscule volume they will be left with by then.
"azfacea singing the death song to hard drives. Again" i didnt know i needed your permission.
cant deal with the argument attack the person. NT
50% drop since 2010. and what happened since 2000 to 2010 was that 50% as well ? u just proved my point, half a node progress in one decade. oh wow hard drives are so alive they are so on fire oh wow
All of the links point to the price of <=4TB drives falling. >4TB prices are higher per TB and there's not much information on whether or not they continue to fall in price in 2019. In particular, try looking at >=10TB models. I've watched them like a hawk in the past year. They're still very very expensive. The deals make them nicer in some cases, but we're talking regular pricing.
Roses are red, violets are blue, we were talking about prices for hard drives not stagnating since 2010 but going down, and your comment that you "very thoroughly" substantiate above does nothing to contradict my argument. I suck at rhymes but I do know how to make a valid point ;).
I've noticed price per TB has finally started to come down as I picked up 16 TB drives compared to my last buy of 8 TB drives. It really was stagnant for a while. I'm curious to see how the pricing for HAMR based HDDs pans out. I won't be jumping onto them because of potential reliability issues (want to wait a few years and see how they do) and I'm sure the price will be pretty bad at the start. Meanwhile, they're kicking out more releases of the existing technology just to soak in the money from releasing slowly instead of just jumping to the greatest capacity that they can do. Ok, they do need to test them as they go and maybe if they go right to the max it'll turn out to be a failure and can't be released.
One worry I have about enormous spinning drives is the ability to back them up. 50 TB at even the actual max speed of 12 Gbps (SAS) would take ~5 days to copy. If a drive fails in a RAID system that's a little unnerving to know it takes that long for the only remaining drive in a single mirror array to copy its contents. I think we're going to enter a time when we have to make several mirrors of data just to keep up with reliability. Thankfully, drives don't die much these days but that goes back to my comment about HAMR and not knowing just how well it's going to hold up.
Ultimately, having high capacity drives helps drive down electrical costs to operate less capacity drives but up goes risk so it's difficult to decide which way is best.
As for consumers using spinning drives, yes, there's a lot less demand for them. Consumers are mostly interested in performance and put their reliance on backup to the clouds these days. But spinning drives are going to be around for a very long time on those servers holding all of this massive data. It's still questionable as to whether SSD will ever overtake spinning drives. We may end up with a new medium which blows both of them away before HDDs are finally retired. SSDs are so fragile that I'm weary of storing important stuff on them.
Have you ever tried backing up this much data reliably? Even Amazon's Glacier is struggling with data retention economically. Ten 9s of durability is nowhere near enough, but a 6-drive Raid 6 (and the equivalent B/Z 2) with HDDs that have unrecoverable read errors every 10^15 bits still has a near 40% chance to fail completely during a rebuild with this much capacity.
At least with ZFS you're likely to only lose one bad sector rather than the whole data set, but that's still a 40% chance to lose at least one piece of data in a rebuild.
With RaidZ2 it would be nearly impossible to lose any data due to URE's.You would have to have a drive fail AND then two more drives fail on the exact same sector. Incredibly unlikely.
I wonder if hard drives have the same aspect of "binning" as CPUs? I know there are factory defect lists, but do they pull the not-so-great platters and specifically put them into lower-capacity drives?
There have been some good innovations in hard drives in recent years, but this is pathetic. Their roadmaps have slipped so many times, they make Intel’s 10nm process look like it was on time. HDD manufacturers have slipped their timelines so much that anyone without a short memory should see this like a go-fund-me/crowd-sourced business plan. As I said originally, this is pathetic. I understand HAMR and MAMR are hard to get going by the fixed costs, but it’s just ridiculous 3-5 years after their promises. Investors should be upset to say the least, especially during an expansion.
The wonderful world of price fixing. You can actually see it happening when Toshiba entered the market and made HDDs that could keep up with Seagates and WDs. A sudden dip in prices. And then, after a few months, they suddenly were all friends and the prices got stable again, even went up again on some models.
Just a simple (maybe dumb) question from avg Joe. Why can't they increase the physical size of HDDs? Why does it HAS to be 3.5"? I know there are standards set for size and the PC Case designs are based on that but those can also be changed.
There are (almost) no dumb questions. There have been many different larger form factors, as discussed briefly at https://en.wikipedia.org/wiki/Hard_disk_drive#Form... This includes the 24-inch IBM 350, which is part of the one-tonne IBM 305 RAMAC computer system (lots of fun reading about these old things), and 14 and 8-inch sizes before more common sizes such as 5.25, 3.5, 2.5 and 1-inch drives. There were consumer products based on the 5.25" form factor, such as the Quantum Bigfoot (https://en.wikipedia.org/wiki/Quantum_Bigfoot), which I remember having in the 90s. With a larger drive you can have more total drive area per actuator (the arm that moves to put the read/write heads in place). This can be good in terms of cost efficiency, but the number of I/O operations per second (IOPS) goes down. I think Seagate is currently working on putting multiple actuators within the same drive, so this partly negates this.
I can think of a few reasons why 5.25" drives are not coming back: - smaller drives mean more drives in the same physical volume, which in turn means more IOPS (many servers moved to 15000rpm 2.5" drives before moving to SSD, which now can have all sorts of exotic form factors like m.2 and long narrow "ruler" designs) - smaller drives means more flexibility in terms of case designs - the main advantage with larger drives would be cost efficiency, but it possible that quality control issues when producing large discs mean more material is discarded due to small defects (but this is likely not at all as bad as in semiconductor manufacturing) - I think there were also some material strength/tension/homogeneity issues with keeping high enough rotational speed at large sizes
In the end it likely comes down to the available market though. Very few people would opt for a physically larger drive when small do the job better but are potentially just slightly more expensive. If there are some hard drive engineers on the forum I'm happy to be corrected if I'm wrong.
Google and others at the OCP started talking about alternate form factors a few years ago, and yes, it's likely to occur. But the cloud storage customers have additional different criteria that doesn't play well for the home user, like reduced error correction in the hardware.
Right now, us data hoarders benefit from the density progress, but I think at some point there will be a divergence and we won't be able to make use of it. Then again, maybe irrelevant. I can't even justify bumping my 6TB NAS drives to the 14s or 16s.
Am I missing something? I thought WD were going to ship MAMR first before moving to HAMR? In terms of DataCentre requirement, even 50TB per HDD is still far too small.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
35 Comments
Back to Article
shabby - Tuesday, November 5, 2019 - link
Can't wait for the 22tb drive in 2021, whose with me?!?azfacea - Tuesday, November 5, 2019 - link
the issue is not max capacity. the issue is TB per dollar. HDD have been quite stagnant since 2010 and will do even worse in 2021 with the miniscule volume they will be left with by then.close - Wednesday, November 6, 2019 - link
azfacea singing the death song to hard drives. Again. About that TB/$, doesn't really show it flatlining as you suggest: https://www.backblaze.com/blog/hard-drive-cost-per...Maybe Backblaze doesn't have access to the same high quality data you have.
Other random links showing a drop which percentage wise is pretty consistent (~50% from 2010, even if in absolute numbers it's low):
https://jcmit.net/diskprice.htm
https://mkomo.com/cost-per-gigabyte-update
azfacea - Wednesday, November 6, 2019 - link
"azfacea singing the death song to hard drives. Again" i didnt know i needed your permission.cant deal with the argument attack the person. NT
50% drop since 2010. and what happened since 2000 to 2010 was that 50% as well ?
u just proved my point, half a node progress in one decade. oh wow hard drives are so alive they are so on fire oh wow
close - Thursday, November 7, 2019 - link
Don't move the goalpost, I thought prices absolutely stagnated since 2010. Now you wanna talk about 2000? Nice way to concede the point ;).ballsystemlord - Wednesday, November 6, 2019 - link
All of the links point to the price of <=4TB drives falling. >4TB prices are higher per TB and there's not much information on whether or not they continue to fall in price in 2019.In particular, try looking at >=10TB models. I've watched them like a hawk in the past year. They're still very very expensive.
The deals make them nicer in some cases, but we're talking regular pricing.
close - Thursday, November 7, 2019 - link
Roses are red, violets are blue, we were talking about prices for hard drives not stagnating since 2010 but going down, and your comment that you "very thoroughly" substantiate above does nothing to contradict my argument. I suck at rhymes but I do know how to make a valid point ;).dudadiesel - Sunday, November 17, 2019 - link
I've noticed price per TB has finally started to come down as I picked up 16 TB drives compared to my last buy of 8 TB drives. It really was stagnant for a while. I'm curious to see how the pricing for HAMR based HDDs pans out. I won't be jumping onto them because of potential reliability issues (want to wait a few years and see how they do) and I'm sure the price will be pretty bad at the start. Meanwhile, they're kicking out more releases of the existing technology just to soak in the money from releasing slowly instead of just jumping to the greatest capacity that they can do. Ok, they do need to test them as they go and maybe if they go right to the max it'll turn out to be a failure and can't be released.One worry I have about enormous spinning drives is the ability to back them up. 50 TB at even the actual max speed of 12 Gbps (SAS) would take ~5 days to copy. If a drive fails in a RAID system that's a little unnerving to know it takes that long for the only remaining drive in a single mirror array to copy its contents. I think we're going to enter a time when we have to make several mirrors of data just to keep up with reliability. Thankfully, drives don't die much these days but that goes back to my comment about HAMR and not knowing just how well it's going to hold up.
Ultimately, having high capacity drives helps drive down electrical costs to operate less capacity drives but up goes risk so it's difficult to decide which way is best.
As for consumers using spinning drives, yes, there's a lot less demand for them. Consumers are mostly interested in performance and put their reliance on backup to the clouds these days. But spinning drives are going to be around for a very long time on those servers holding all of this massive data. It's still questionable as to whether SSD will ever overtake spinning drives. We may end up with a new medium which blows both of them away before HDDs are finally retired. SSDs are so fragile that I'm weary of storing important stuff on them.
trparky - Tuesday, November 5, 2019 - link
You know what they say about large hard drives right? The larger they are the longer you're going to cry when it dies and it takes your data with it.Flunk - Tuesday, November 5, 2019 - link
It's a shame that backing up data is impossible.trparky - Tuesday, November 5, 2019 - link
It's not impossible however it is expensive. At those capacities getting multiple drives for a RAID array can get quite expensive.UpSpin - Tuesday, November 5, 2019 - link
Luckily RAID is not a backup, so it won't be quite expensive.khanikun - Wednesday, November 6, 2019 - link
LTO-8 Backup tapes and drive. 12 TB native or 30 TB compressed per tape. You're still looking at like $4000 for the drive and $150 for each tape.For home users, we just RAID it together and pray we never lose multiple drives as once. Hope & RAID is our "backup" solution.
yetanotherhuman - Wednesday, November 6, 2019 - link
I personally buy 3 drives every time I want one.. one for my PC, one to expand my backup, one to expand my off-site, air-gapped backup.patrickjp93 - Wednesday, November 6, 2019 - link
Have you ever tried backing up this much data reliably? Even Amazon's Glacier is struggling with data retention economically. Ten 9s of durability is nowhere near enough, but a 6-drive Raid 6 (and the equivalent B/Z 2) with HDDs that have unrecoverable read errors every 10^15 bits still has a near 40% chance to fail completely during a rebuild with this much capacity.https://www.wintelguy.com/raidmttdl.pl
At least with ZFS you're likely to only lose one bad sector rather than the whole data set, but that's still a 40% chance to lose at least one piece of data in a rebuild.
extide - Wednesday, November 6, 2019 - link
With RaidZ2 it would be nearly impossible to lose any data due to URE's.You would have to have a drive fail AND then two more drives fail on the exact same sector. Incredibly unlikely.danielfranklin - Tuesday, November 5, 2019 - link
Bring back to 20MB HDD!Scott_T - Tuesday, November 5, 2019 - link
so you can play space quest without having to switch floppysGreenReaper - Tuesday, November 5, 2019 - link
I wonder if hard drives have the same aspect of "binning" as CPUs? I know there are factory defect lists, but do they pull the not-so-great platters and specifically put them into lower-capacity drives?shabby - Tuesday, November 5, 2019 - link
Yes... they put them into the consumer drives 😂RedGreenBlue - Wednesday, November 6, 2019 - link
There have been some good innovations in hard drives in recent years, but this is pathetic. Their roadmaps have slipped so many times, they make Intel’s 10nm process look like it was on time. HDD manufacturers have slipped their timelines so much that anyone without a short memory should see this like a go-fund-me/crowd-sourced business plan. As I said originally, this is pathetic. I understand HAMR and MAMR are hard to get going by the fixed costs, but it’s just ridiculous 3-5 years after their promises. Investors should be upset to say the least, especially during an expansion.Beaver M. - Wednesday, November 6, 2019 - link
The wonderful world of price fixing.You can actually see it happening when Toshiba entered the market and made HDDs that could keep up with Seagates and WDs. A sudden dip in prices. And then, after a few months, they suddenly were all friends and the prices got stable again, even went up again on some models.
TheOtherOn - Wednesday, November 6, 2019 - link
Just a simple (maybe dumb) question from avg Joe. Why can't they increase the physical size of HDDs? Why does it HAS to be 3.5"? I know there are standards set for size and the PC Case designs are based on that but those can also be changed.AdditionalPylons - Wednesday, November 6, 2019 - link
There are (almost) no dumb questions. There have been many different larger form factors, as discussed briefly at https://en.wikipedia.org/wiki/Hard_disk_drive#Form...This includes the 24-inch IBM 350, which is part of the one-tonne IBM 305 RAMAC computer system (lots of fun reading about these old things), and 14 and 8-inch sizes before more common sizes such as 5.25, 3.5, 2.5 and 1-inch drives.
There were consumer products based on the 5.25" form factor, such as the Quantum Bigfoot (https://en.wikipedia.org/wiki/Quantum_Bigfoot), which I remember having in the 90s.
With a larger drive you can have more total drive area per actuator (the arm that moves to put the read/write heads in place). This can be good in terms of cost efficiency, but the number of I/O operations per second (IOPS) goes down.
I think Seagate is currently working on putting multiple actuators within the same drive, so this partly negates this.
I can think of a few reasons why 5.25" drives are not coming back:
- smaller drives mean more drives in the same physical volume, which in turn means more IOPS (many servers moved to 15000rpm 2.5" drives before moving to SSD, which now can have all sorts of exotic form factors like m.2 and long narrow "ruler" designs)
- smaller drives means more flexibility in terms of case designs
- the main advantage with larger drives would be cost efficiency, but it possible that quality control issues when producing large discs mean more material is discarded due to small defects (but this is likely not at all as bad as in semiconductor manufacturing)
- I think there were also some material strength/tension/homogeneity issues with keeping high enough rotational speed at large sizes
In the end it likely comes down to the available market though. Very few people would opt for a physically larger drive when small do the job better but are potentially just slightly more expensive.
If there are some hard drive engineers on the forum I'm happy to be corrected if I'm wrong.
bsd228 - Wednesday, November 6, 2019 - link
Google and others at the OCP started talking about alternate form factors a few years ago, and yes, it's likely to occur. But the cloud storage customers have additional different criteria that doesn't play well for the home user, like reduced error correction in the hardware.Right now, us data hoarders benefit from the density progress, but I think at some point there will be a divergence and we won't be able to make use of it. Then again, maybe irrelevant. I can't even justify bumping my 6TB NAS drives to the 14s or 16s.
Beaver M. - Thursday, November 7, 2019 - link
Vibration, speed, reliability.Look at the WD Raptors. Why did they put 2.5" HDDs in a 3.5" case? And then you have the answer.
SirCanealot - Wednesday, November 6, 2019 - link
Be great once these are available for a consumer to buy at a reasonable price... in 10 years or so :/I got a reasonably-priced 4tb drive a while back, but over this it starts to get way too expensive imo.
ksec - Thursday, November 7, 2019 - link
Am I missing something? I thought WD were going to ship MAMR first before moving to HAMR? In terms of DataCentre requirement, even 50TB per HDD is still far too small.Lakados - Thursday, November 7, 2019 - link
DataCenters are moving to a custom size format, the 2.5" and 3.5" format is too limiting.Samus - Friday, November 8, 2019 - link
So much for MAMR in 2019 like WD promised. They're just catching up to Seagate now, and I'm no Seagate fan so that's unfortunate.sanwizard - Wednesday, February 12, 2020 - link
Mostly consumer viewpoints posted here. As stated in the article, these drives are intended mainly for enterprise/cloud scale use cases.As an example, Infinidat storage systems can rebuild a failed 12tb drive in about 30 minutes.