Trouble at t'spinning rust mill: Disk drive production is about to head south A sea change is looming in the disk drive industry as drive spindle motor maker Nidec reduces its total disk drive demand number. Nidec is a Japanese company and has the largest market share of any such manufacturer. It foresees demand for about 400 million units in 2016, down 13 per cent from its previous October 2015 …
This post has been deleted by its author
I mean, the "need" to "consume" more data is growing fairly... steeply, istn't it? - regardless of how poor the sales of PCs are. People WANT a lot, lot more space for their data, and won't send their terabytes to the cloud, and won't spend that much for very large ssd either, because the prices are still way too high to make ssd a popular storage format (for the public at least).
Does your average end-consumer really have terabytes of data? The only possible need for that kind of space most people have would be storing massive media files, and few people do that. They stream.
Beyond that they have photos and other small files that their present setups are perfectly capable of handling. They already have HDs that are under-utilized. They do not need to buy massive new drives. And new consumer devices are more and more likely to be SSD of smaller sizes.
And on the other hand, companies are moving towards cloud storage and/or SSD. The days of millions of little spinning rust wheels are numbered.
From the consumer perspective, the move is from PCs that have 'proper' hard drives to tablets & phones that have some form of SSD storage backed up by some form of cloud storage. The days of having 10 times as much storage as you need are over.
From the cloud side, there may be more data being stored, but there's much less wasted space as you're no longer using 'your' hard drive, just a proportion of some sort of drive pool.
Be interesting to know the efficiencies in cloud due to dedup - anyone got any stats?
It's generally understood dedupe can be very useful in VDI, which is not really a static environment. On the other hand video archive is as static as it comes but only benefits from dedupe if completely unmanaged. As the folk at DD and elsewhere have shown dedupe is a very useful technology for backup. Compression tends to be much more effective than dedupe in the database world. Ultimately it's the nature of the data set that determines if dedupe may be useful and while plenty have over sold it that doesn't mean it, and its often forgotten twin compression, are not useful across a wide range of workflows.
At the office level companies are no longer replacing their PC every three years, so every PC not sold is a hard disk not sold. And if they have any sense, when they do eventually replace the PCs they will specify a new PC with a smallish SSD to boot from (much faster and more reliable for read-mostly usage, no longer an expensive option). People in offices should (mostly) be storing the user data via the LAN to a server, so a boot drive should be all that is needed.
It'll be a long slow decline (which won't necessarily lead to extinction). People recording video (PVRs and Security systems) will happily fill tens of Tb, if it's cheap enough. Then there are the corporate backup systems, and CERN, and the three-letter agencies....
"How can they [stream] with such tight data allowances?"
Obviously most people's data allowances aren't all that tight. BT only have limited data allowances on their absolute cheapest packages - anything over £10/month is unlimited - while Virgin don't appear to offer anything other than unlimited. Phones tend to be tighter, but then streaming is likely to be done over wi-fi anyway. Apparently things are a lot worse in the US, but on this side of the pond data allowances just aren't generally a big problem.
"Furthermore, can they really trust the cloud to always be there?"
They can and do. Whether they should is an entirely different question, and one the vast majority of the general public will never even consider.
You are correct - the average end-consumer doesn't need terabytes of data on a home pc. Most people don't need a home pc if they only browse the internet and read email. They (correctly) assume their emails will always be there, stored in the cloud.
Photos are almost all taken on their phones, which also 'magically' store the pictures somewhere else, just not on their SSD phones.
Yes, some companies are running their businesses wholly from AWS et. al. cloud services. Now the big question - just how do you think Facebook, Google, Twitter, Amazon, etc are storing all this data? Does it have to be always-on, always at near-memory speed access? Do you really think billions of my-trip-to-the grand canyon/cat pictures taken 10 years ago and not looked at in the last 5 years need to be instant-access SSD stored? Your miles of business email accounts from employees long gone. Yes, they need to be saved but not on flash storage. There's a need for speed and there's a need to keep this stuff for a long time. Spinning rust is good enough for the latter.
it's true that fewer and fewer people use digicams, like our family, but people do snap a lot, and the instinct is to cling to your milions of i-shots, regardless of their value (ANY value!). The volume is creeping into terabytes country, and then there's video people do record in ever higher resolutions.
People don't record much video, because taking photos that look sort of about the same category of as good as professional photos isn't terribly hard, but doing that with video is really quite challenging.
The restriction on numbers of photos taken, even with good DSLRs, stopped being SD-card capacity a couple of years back - I'm a pretty enthusiastic photographer, took three 32G SD cards on a three-week trip to South-East Asia and only used half of one. People realize that they don't bother checking through the nineteen shots of a badly-posed monkey that they took by holding the button down, and move to pressing the button once.
Does your average end-consumer really have terabytes of data?
Probably not, but for the rest of us, unfortunately, lower demand for disk drives means higher cost.
I may have to move up my HDD replacement schedule and buy both 5TB drives this year.
1TB SSD + 10TB HDD (WD Black) = sufficient storage for now. Mind you, I had a nightmare where I was trying to back it up to optical disks
The data is accumulated on non-PC devices nowdays and from there all usual suspects do everything in their power to move the data to "their" cloud and not to consumer's own storage.
Starting with small annoyances like Android removing mounted pass-through SD card support and finishing with outright sabotage.
My ST238 sees your ST225 and raises you 6%.
... My God, what we used to do for a little extra capacity. RLL, what a disaster. Never saw one that was worth the powder to blow it to Hell. I still have an ESDI drive and controller in the garage, but I don't think they work any more. Full height 5 1/4 inch and weighs a ton.
The capacity of every one of those drives I've ever owned added together is barely a footnote, less than the bad sector table of any modern drive.
" I wonder what sort of capacity you'd get if they made them with today's platter density."
You wouldn't. The reason they weren't reliable was due to the platter size.
inner linear velocity vs outer linear velocity varies so much that it affects the fly height. That wasn't a problem 20 years ago but fly height now is 1/1/0 of what it used to be.
Large platters need to be much more rigid (and therefore heavier), or (you guessed it) it affects the fly height), which means more startup torque, which means more current draw.
The arms need to swing through a much wider arc.
etc etc etc.
5.25 inch drives are gone. They're not coming back.
There was nothing wrong with RLL .... unless you were doing it with an ST-238! What a disaster those drives were. Worst hard drive ever ... well ... apart from the .... and of course the old ..... No. Let's not go there. We'd be here all night and it's time for my slippers and milk arrowroot biscuit. Goodnight nurse.
"High-capacity (nearline) enterprise HDD shipments are now estimated to grow from 37 million units in 2015 to 48 million units by 2020."
Pardon the pun but it's all about how you spin it. With HDDs set to on average get significantly bigger and increase in number this report is suggesting that disk drives will remain a very significant part of many companies storage plans for at least the next five years. Winter is coming for HDDs, but for shared storage it is still some distance off. Be careful who you back in the storage Game of Thrones, after all Dell buying EMC is like House Frey taking over House Lannister (and you can be sure there will be storage blood spilt at the Dell | EMC Blue Wedding). No one can be sure what will happen next...
au contraire to au contraire.
There are plenty of examples in the vdeo where the glottal stop is used as I recall it from my time spent living in the north of England. I would say that the occasions it is not used are when the performers are forcing the words to fit the rhythm of the music.
I think the author of this story hit the lede: that analysts who presumably know what they're doing have reckoned that there won't be an explosion in data-storage demand from the cloud companies, that they'll keep buying discs at the rate they have been rather than continuing to grow demand at 20% a year.
"Big Data Won't Be That Big" seems a perfectly registerian headline ...
Corrupted multinationals managers. That is the real issues.
And the myth that no one would be buying Hard Disk Drives anymore... the biggest lie ever.
Only Intel/Micron 3D XPoint could in the future kill the Hard Disk Drive. But that isn't going to happen anytime soon anyway. For sure 3D XPoint is going to kill the unreliable NAND SSD tech.
Have a 1TB SSD. About to buy another. For a damn laptop.
Sorry, but the price is well worth it and even though I remember paying £1000's for a 40Mb hard drive and 2Mb RAM, I rarely spend on anything IT.
But I bought a gaming laptop. It does everything a normal laptop can do, plus play every game I own, and I don't have to manage it or upgrade it at all (several years old, laughed at GTA V). Into that I put a 1Tb SSD as a treat to myself:
- Data Size: Good
- Phsyical size: Smaller, lighter and thinner than the 2,5" 1TB HDD it replaced
- Speed: Phenomenal
- Reliability: Impeccable so far and I'm much happier with taking the laptop in the car/plane with the jolting now there's only one moving part in it (a fan, which incidentally is the ONLY thing I've replaced in my laptop in its lifetime, because it broke and couldn't cool effectively).
- Cost: On the order of twice the equivalent HDD. To be honest, the speed alone is worth that.
On an un-upgradeable machine, it feels like a new lease of life and a massive upgrade. Programs load, switch and close faster. Boots are on the order of seconds even with a HUGE user profile. Games load as fast as they can get into the graphics RAM. Alt-Tab is now viable to check a webpage and then switch back to a live game and you don't even feel it. And games that chopped on loading edges (e.g. GTA V, Batman, etc.) no longer do. I've noticed more difference putting in an SSD than I did doubling RAM on previous machines or moving between two laptops - one several years old and one brand-new at the time.
HDD is dead. Costs will come down. Focus will shift from HDD to SSD (it already is) and then its game over. HDD can't compete in terms of speed. HDD can't take a jolt while its active. And HDD isn't that much bigger at the moment than SSD in a similar form factor (the fact that almost all SSDs are 2.5 inch kind of tells you they could meet capacity at any moment, it's just the cost that's the factor.
My laptop has two drive bays. I deliberately kept the old HDD in the other in case something went wrong with the SSD. A year down the line, I'm just going to buy another 1Tb 850 EVO and shove it in. To be honest, I'm eyeing up the 2Tb.
In work, I have a choice this summer upgrading all clients to double the RAM or putting in SSDs. I'm seriously considering both, but I'm also seriously considering just the SSD if it comes to it. A tiny drive is all that's needed for client machines and makes a huge difference to perceived speed, even if the network will always be the bottleneck.
Stop messing about and move all that HDD manufacturing to SSD. There's nowhere you CAN'T use SSD. You could put it in cars or mobile devices or laptops or servers or storage arrays. There's quite a few places you can't use HDD, though. And the price difference is really not that much on the "sweet spot" drives (where cost per Gb actually makes sense for most people who would buy their own drives).
"Stop messing about and move all that HDD manufacturing to SSD. There's nowhere you CAN'T use SSD. You could put it in cars or mobile devices or laptops or servers or storage arrays. There's quite a few places you can't use HDD, though. And the price difference is really not that much on the "sweet spot" drives (where cost per Gb actually makes sense for most people who would buy their own drives)."
Oh? What about areas where capacity trumps speed like media serving and video surveillance? Then the price premium doesn't make sense since in media it's the network that's the bottleneck nine times out of ten while you really want to store everything in one place for easy reach (and just keep a mirror in case of a failure). Meanwhile, surveillance has a predictable data transfer rate and really only needs capacity to store more footage for security purposes. In both cases, your data needs may easily run into the 4 or 5TB range, and given SSD drives of that size easily run four figures while rust can handle it in the low threes...
DVR? Almost entirely 3.5" drives. This is my point. Almost all SSD are 2.5" and virtually empty inside at the moment. The cost comes down and you can cram DOZENS of those Tb into one 3.5" casing. Get the cost down, you can make a drive large enough to compete with anything HDD could ever do.
And it'll support multiple full HD streams writing without any effort at all - at the moment you can't guarantee that and need to have some hefty compression on high-end chips to make it viable if you have a lot of channels, and there are the range of always-on drives (like WD Purple? I think) which are designed especially for that and come with a premium too.
Again, your objection is cost. And cost comes down with volume. And volume increases based on making all the low-hanging fruit SSD. So although DVR isn't the immediate first choice, it will benefit from every PC coming with - as they now do - a choice between a 1Tb hard drive or a 256Gb SSD.
"Again, your objection is cost. And cost comes down with volume. And volume increases based on making all the low-hanging fruit SSD. So although DVR isn't the immediate first choice, it will benefit from every PC coming with - as they now do - a choice between a 1Tb hard drive or a 256Gb SSD."
If you're only talking about clients (a miniscule part of the disk market), then sure. If you're talking about the vast majority of data storage requirements in the real world (i.e., the JBODs which all DCs really run on), then no.
Problem with the 'economy of scale' argument, and the reason why it's proven to be ineffective so far, is the initial capital cost of flash fabs are ridiculous. Really, really ridiculous. See, while you can churn out a HDD plant in Indonesia for <$1bil and start mass-producing 10TB drives, building a flash fab costs 15-20 times as much... to produce a product which, as you note, retails for only 2-3 times the price. You're looking at a ten year runway just to clear the setup costs, let alone running costs. The economics don't work, and everyone in the storage industry knows that they don't; half the flash fabs in the world have yet to recoup their startup cost. To produce enough SSDs to just reach 50% of needed storage output, you'd need half a trillion dollars worth of new fabs every 4-5 years. Odds are, they will now turn a profit before the tech itself is obsolete.
That's why flash output simply isn't increasing at the same rate as data storage needs are; no-one wants to throw 15 billion dollars at a tech which is under threat from 3-4 superior competitors that may be market-ready by 2020. Flash growth rates continue to lag behind data growth rates, and plain old HDDs are the ones which are taking up the slack. Flash is pretty much already dominant in the areas where it's going to dominate (high-performance workloads that we used to stick on the now-moribund 15000rpm market and the upper-end 12000rpms) and is making no real impact on the serious HDD markets (7200 long-term storage and mid-level access jobs). Serious storage guys measure workloads in petabytes, not TB or GB, and they simply don't talk about flash because the amount of it in production use is fractions of a percent.
The story of this article is just a spindle maker saying they're going to sell less spindles; that doesn't signal that people are moving to Flash but simply that individual HDDs are getting bigger, and so will sell for a correspondingly higher margin. That's all.
I'm shocked it was still going up in 2015! I would have guessed this happened a few years back, given the rise of SSDs and the large decline in the PC market since its peak five years ago.
I assumed the article was going to say that the number of TB shipped in hard drives had dropped, but I guess if units are only falling now, it will take a few more years for the total TB to decline.
"Damn, I'd thought I'd seen niche products"
HDD motors are only one of the types Nidec make. The internals of a huge number of computer fans are nidec-cored and you'll find their products embedded in a lot of other technologies.
It's a niche, but they're far from a niche manufacturer.
I currently have my 1TB Laptop HDD at 91% capacity. I might be able to get a 2TB HDD at most, but not more. Why? Because nobody seems to care enough to get 4TB or even 3TB HDDs made in the traditional laptop form factor. Never mind that the PS4 also uses this format, and 4TB is something very useful there. This is because many people are buying into the SSD craze. I would do so, but 1TB is still too expensive and I would still be stuck with the same issue I currently have. I haven't seen any 2TB SSDs.
On desktop systems, I'm happy with Seagate still churning out higher capacity HDDs. Hopefully they'll remain doing so, as large storage is still a need for me. I'm planning on upgrading my "home server" PC, which currently has 6.5TB storage capacity.
"I haven't seen any 2TB SSDs."
http://www.samsung.com/us/computer/memory-storage/MZ-75E2T0B/AM
http://www.samsung.com/us/computer/memory-storage/MZ-7KE2T0BW
http://www.samsung.com/us/search/searchMain?Dy=1&Nty=1&Ntt=pm863
Now you have (and 4TB).
The 863s are pricey but SSD prices are still falling every few months.
Novachips are flogging 8TB drives too: http://www.theregister.co.uk/2015/03/05/eight_terabyte_ssd/
If you have $13k burning a hole in your pocket you can even buy 13TB: http://www.pcworld.com/article/3021886/hardware/the-worlds-first-13tb-ssd-is-here.html
The drop in demand for consumer Winchester drives isn't about at switch to SSD. Punters are moving away from PC's containing drives, period. So a drive manufacturer having a SSD products isn't really going to help.
An SSD is a fudge - it gives Flash ROM an IDE interface, so it's plug-and-play with MS-DOS and Windoze. IDE was also a fudge to make poor-man's SCSI look like an ST506.
If the OS supports Flash AS Flash for backing store, packaging it up to look like a discrete IDE-compatible drive is a waste of time. If WDC hopes to survive by doing this, they're going to be out-of-luck - manufacturers will start buying their flash in the same format they buy their RAM.
That said, as a tech writer I learnt very quickly that predicting the demise of the winchester was risky. They've seen off bubble memory, serial ROM, wafer scale integration... When it comes to $/Mb (or Gb/Tb), they're always ahead.
But when it comes to 'price for the smallest thing big enough to do the job', SSD is already winning; the smallest SSD you can get is big enough, and is cheaper than the smallest HDD you can get.
There were a couple of years when I had compute nodes with the OS on 8GB USB2 sticks; I don't recommend this, they do wear out and reinstalling the OS even on a compute node is never as painless as you hope (CUDA drivers :().
Not just Windows but also with old motherboards that were built before the idea of flash as a separate interface even existed. Boards where the main reason for PCI Express was graphics and thus only had one x16 slot and maybe one or two x1 slots that may well be blocked by a double-height graphics card. Trust me, a Core 2 Quad (especially the 9000-series) still has plenty of legs so those machines still need to be accommodated.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
