Hmm, would like to know what these tighter tolerances will do for their reliability, though. Plus as 3D Flash ramps up, there's still the chance of a 16TB rust drive getting matched in silicon.
Spinning rust fans reckon we'll have 18TB disk drives in two years
We could see nine platter helium-filled disk drives because the manufacturers can cram more thinner platters inside a helium-filled enclosure. According to a TrendsFocus report, current disk platter separator components could be eliminated and enable eight or nine platters to be used in the same 3.5-inch enclosures that …
-
-
-
Monday 11th April 2016 12:44 GMT Mage
Flash are not all equal.
Cheaper ones have poor number of write cycles.
Expensive Flash powered down may not last as long as powered down non-shingled HDD
Cheaper ones beat HDD on random access, but can MUCH slower for sequential writes, or even random writes.
"Flash" covers a very wide variety of qualities.
-
-
-
-
Monday 11th April 2016 18:05 GMT Anonymous Coward
Re: Wishful thinking ...
If they wanted to they could offer 100 TB SSDs, the fact those don't exist is due to a lack of market, not because they can't make them.
But as pointed out, what matters is getting the cost per TB reasonably close. I think when SSDs are less than twice as much is when spinning media will disappear entirely. We're a long way from that day. But we're close to the point where it doesn't make sense for home users to have anything but SSDs even for bulk storage. Unless you collect stuff off bittorrent or have a child you video constantly, you aren't needing 8 TB drives today and 14 TB drives tomorrow.
-
Monday 11th April 2016 18:21 GMT Steven Raith
Re: Wishful thinking ...
DougS makes a valid point.
If you're running a datacentre with big storage needs, just the cost/rack savings alone of switching from 2tb spinning disks to 16tb SSDs would pay for themselves not only in real estate, but power usage (don't forget the host chassis you can power down/repurpose) and speed, meaning you can get more work done - and work = money.
At home...well, a 16tb SSD would be nice, but I don't need it at the moment. When it gets to within about 125% the price of spinning media, it'll look more tempting.
But yeah, traditional HDDs are starting to look like they are in a dicey position lately.
Steven R
-
-
-
-
-
-
-
Monday 11th April 2016 12:56 GMT JRW
Re: Question
Why No Super-Sized (physically) Hard Drives?
Older hard drives used to be both taller and wider. Current large drives are 3.5 inch and as someone has said can fit a slot in a 1U server (though most servers go for 2.5 inch these days). I can remember 5 and 8 inch drives and they were taller as well as wider. Before these there were even larger drives, mostly so expensive people still used paper tape/cards as storage. I was told it's a bit like silicon wafers - the smaller the unit the better yields so lower costs. The larger formats were dropped because the smaller ones made more sense. Down to the 1 inch drives which were killed by falling flash prices and the cost of engineering moving parts that small. Current price ratio of flash to large drive is probably a little under 10:1 and in slow decline, so drives will be with us for a while. Since the platters have been aluminium or glass for some years the risk now of rust being an issue is probably very low :-)
A brief history of hard drives is at: https://en.wikipedia.org/wiki/History_of_hard_disk_drives Warning, it's not very exciting.
-
Monday 11th April 2016 19:02 GMT Alan Brown
Re: Question
"Current price ratio of flash to large drive is probably a little under 10:1"
It's about 4:1 on consumer drives and 2.3-3:1 on enterprise drives (or less).
This is mainly because enterprise HDDs can be silly-expensive whilst the enterprise versions of most SSDs are only a few percentage points more than consumer ones.
(One example: 2TB Hitachi HDDs are about 90 quid - or 350 if you want one certified for a Nexsan Satabeast array.)
-
-
Monday 11th April 2016 12:34 GMT Jon 37
Re: Question
Because economies of scale, and chicken/egg. (And partly because 1U servers are a thing and what you're describing wouldn't fit in a 1U server).
If you could wave a magic wand and have loads of servers with double-height bays and loads of double-height drives available, then they'd have economies of scale, and there would be a market with multiple suppliers and multiple consumers. Unfortunately, there's no magic wand. So someone has to be the first drive manufacturer to produce this kind of drive, which has no market. Then some server vendor has to invest in producing suitable servers. Due to these up-front costs, and the small market meaning no economies of scale, prices will be higher. That makes your new format expensive, so it's unlikely to take off, so it probably won't survive. So the drive manufacturer and server vendor are likely to lose their investments, so no sane drive manufacturer will make such a high-risk investment in the first place.
However, there may be a way around that: Google. Google has suggested taller drives, and Google buy enough hard drives that *just selling the drives to Google* the drive manufacturer would still get some economies of scale. Google also make their own custom servers already, so they can decide to include these new bigger hard drives. Then the drive manufacturer(s) can then try to sell to the new drives to other cloud vendors, and maybe in time the drives will trickle down to "normal" (non-cloud) servers.
See http://www.theregister.co.uk/2016/02/26/google_cloud_disk_white_paper/
-
-
Monday 11th April 2016 13:11 GMT JRW
Re: Question
They would be expensive as low volume and the rebuild times would be unbelievable high. Basically no one woudl buy them. Let's say these devices can sustain writing at 150MB/s. It would take 11+ hours to fill up a replacement 6TB drive if it were full (assuming no other bottlenecks). On some older designed storage platform a rebuild of a multi-TB drive can be weeks. Dual parity is looking inadequate to protect data on current large drives, even on well designed platform. Your 5 inch drives would be an order of magnitude worse. With large drives (and flash devices) the need is not to step up to physically larger devices but for better protection systems than dual parity. The challenges for storage are not how to get the most data on one device but how to (a) protect the data and (b) get the performance required at a price people can pay. If the data requires so little performance/access it would work in a 200TB drive then a tape will be cheaper.
-
Monday 11th April 2016 19:03 GMT Alan Brown
Re: Question
"Dual parity is looking inadequate to protect data on current large drives,"
Not "looking", just "is"
We lost several raid6 arrays for various reasons over the last decade. This was one of the prime motivating factors to push onto RAIDz3, as recovery from backups is time consuming.
-
-
Monday 11th April 2016 18:14 GMT Sgt_Oddball
Re: Question
I'd think it was practical limitations that would cause issues. I mean a 3.5 inch disk (platter around 3.4 as a back of a fag packet guessitmation to account for the walls of the disk) travels at about 150 mph... Then a 5.25 disk will hit about 230 mph of both clocked 15k rpm for the fastest disks, or a 70 mph increase which would increase the likelihood of catastrophic failure a fair amount I'd wager...
-
Monday 11th April 2016 19:17 GMT Alan Brown
Re: Question
"Then a 5.25 disk will hit about 230 mph of both clocked 15k rpm for the fastest disks, or a 70 mph increase which would increase the likelihood of catastrophic failure a fair amount I'd wager..."
Kinetic energy goes up with the square of the velocity - and the area under stress goes up with the square of the radius.
I'm not taking your bet.
-
-
-
Monday 11th April 2016 19:31 GMT Michael Sanders
Everyone is mostly wrong regarding: Question
It has to do with the motor's ability to rotate more/larger discs. 3.5" size and 3 platters is really the optimum. 4 or more do exist but not at 7200 RPM or faster. But yeah, I do miss the days of full and double height drives...and 5.25 drives. I had this same question a few years back about bringing back the Quantum Bigfoot. Even though that drive was a pile...
-
-
This post has been deleted by its author
-
Monday 11th April 2016 16:14 GMT Anonymous Coward
Re: Still waiting....
re. pricing, while you're right about the post/pre flood prices, you CAN get decent 4 TB hdd for around 100 quid. I mean, if you trust WD, you can get one of theirs. If you trust Seagate, you can get a 5 TB for around 90 quid. Do I trust either? No, but what choice do I have? They all get crappier and crappier these days (unless you pay for the upper-spec ones), although it seems that WD are about 10 - 20% pricier (looking at the 4/5 TB external offerings).
-
-
Tuesday 12th April 2016 04:20 GMT I Like Heckling
Re: Still waiting....
I have 2 PCI-E slots on this system... one is covered by the double width video card and the other contains a USB3 PCI-E card (because this silly motherboard only had 2 rear slots and no onboard header) to enable the use of the front slots on the silverstone chassis.
I do have a PCI SATA card in a drawer that would fit in the case... theoretically that would allow me to plug the BD drive into it... But the problem of hard drive overpricing still remains.
-
-
-
Monday 11th April 2016 15:27 GMT Lee D
No good if the speeds don't increase to match. All you've done is increase the backup times by 18 x that of a 1Tb drive.
And with SSDs fast catching up (inconcievable 5-10 years ago, and now they are being offered as options on ordinary mail-order Dell machines, etc.) HDD is going to struggle to compete as ONLY size is their winning stat at the moment. Heat, power, vibration,speed, etc. they all lose at and reliability is about even.
I'd much rather we forgot about all this fast-spinning helium stuff and bought more SSD to bring the cost of their parts down and their reliability up (even if that's only by playing the numbers).
-
-
Tuesday 12th April 2016 09:38 GMT Anonymous Coward
Never mind the backup, it's the restore that matters
To reply to Lusty someone once said to me "I've done loads of successful backups, just never a successful restore." Incremental backup techniques shortened backup windows and reduced the performance needed for backup years ago. BUT backups are not where things are critical, it's the restore, and here multiple drives has a massive advantage over a single massive drive or tape. Some restores are OK taking days, but don't try standing in front of the CEO to tell him a critical system is off line and the restore will take another 2 days. Sadly backup/restore is the most neglected are of IT.
-
-
-
Monday 11th April 2016 15:54 GMT Jim O'Reilly
All wishful thinking!
Large size drives won't work. Tolerances won't allow the bit densities we get with 3.5 or 2.5 disks. Even if we had them, there are no servers that can accomodate them. Server density would be half what it is today, too, while drive power would be more like 30 to 40 Watts.
Tall stacks are more feasible, but tolerances are again an issue. Those actuator head arms can't get much thinner, and that's a real limitation on how many platters are possible. We might see one more platter in a standard full-height drive. Bit densities are at their limit, while going to HAMR will probably mean one fewer platter because the head structure is bigger.
With 16TB SSDs already announced, with SSDs cheaper than enterprise hard drives (at least in distribution and from Dell!) and with Google stating that MLC drives wear as well as SLC drives in real life, spinning rust is in a battle for relevance. Prices are dropping fast and even bulk storage will be challenged next year by 4-bit-per-cell technology and 3D NAND. Oh, and I forgot to mention that SSDs are 100x faster!
-
-
Monday 11th April 2016 18:12 GMT zemerick
Re: All wishful thinking!
It depends on the size. At 500GB it's under 3:1 ( unless you want to test your luck with a bargain bin refurb, which is still only about 4:1. )
SSDs have been plummeting in price, doubling size every year, all while HDDs have been nearly stagnant.
I'm personally done with HDD. SSD are big enough and cheap enough to make the switch now.
SSDs are moving so fast, I have to look it up every single time I talk about it. I have a friend looking to get a 1tb ssd in a few months, and was wondering what it would cost. I started saying probably a bit over $500, then caught myself. A few months ago I bought an 850 EVO 500GB for $147. Even the 850 Pro is under $500 for 1TB, while a Mushkin is only $210.
-
-
Monday 11th April 2016 18:24 GMT Anonymous Coward
Re: All wishful thinking!
Wishful thinking indeed. Ebuyer. 1TB. Cheapest HDD £37. Cheapest 1TB SSD £248. 6.7:1 ratio.
Cheapest 2TB SSD £557. Cheapest 2TB HDD £62. 8.9:1 ratio.
Flash is coming, For performance it already rules, but for capacity that doesn't need lightening speed HDDs will be with us for a few years yet.
-
Monday 11th April 2016 19:40 GMT Michael Sanders
Do you trust?
Shingled drives are wringing out the shammy/caching out the bowl just a little too much. Solid State is pretty much king. It's ridiculous to think its not getting cheaper faster. One only has to look at my own upgrade experience to see that. 220Gig last year/480Gig this year, -same price-
The only thing Magnetic storage is good for now and going forward is for backup and for data sensitive servers where speed is not a factor. It will stay the medium I use at home for my server for example.
I would not trust my data to an overlapping write scheme. And a helium drive needs to show me a significant capacity difference and incorporate a sensor that lets me know when it's leaking out.
-
Sunday 17th April 2016 16:18 GMT Alan Brown
Re: Do you trust?
" And a helium drive needs to show me a significant capacity difference and incorporate a sensor that lets me know when it's leaking out."
Helium drives aren't pressurised, which means that it's just gas diffusion through the (metal) seals to worry about. (Same issue with not being under vacuum, although in that case it tends to be nitrogen diffusing _in_ which causes problems, especially in HeNe lasers)
It's highly unlikely that a helium-filled drive will lose enough gas over its 5 year lifetime to matter and even if it does you'll have plenty of warning as the platter motor current will rise with the increased operational drag.
WRT helium for drives vs balloons, if you work out the volume inside a hard drive vs the volume of a party balloon (and take into account that it's slightly (1-4%) over 1 atmosphere inside the balloon), you'll quickly realise that the gas from one balloon would easily fill 20-40 hard drives (if not more).
That said - I won't buy helium-filled drives for my data centre. The cost differential isn't worth the extra storage space and before anyone bangs on about operational costs, the power consumption of SSDs is small enough that it doesn't need those to drop in price by much to justify purchase on TCO grounds (and SSDs are almost entirely in 2.5" format, so you can pack in a lot higher density in a disk drawer if that's what you want to achieve).
The economics might change for a mostly-powered-down MAID setup but that's not the kind of data storage I have to deal with. Spinup delays would kill that advantage but I'd be more willing to consider MAID for SSD arrays (FWIW measurements here show an idle 4TB PM863 draws next to nothing, so do you even need to consider "power saving" modes?)
-
-
Monday 11th April 2016 21:11 GMT John Savard
Bit-Patterned Media
I hadn't been keeping up with my buzzwords, and thought that shingled recording was a term for what actually is known as "bit-patterned media". Instead, shingled recording involves having the tracks overlap - stopping every few tracks for one that is not overlapped, so that the tracks in a group can always be written in the same direction.
And they now make drives that still do the work of managing this internally, but with special driver software allowing the layout to be optimized depending on which files are accessed sequentially, and which ones are accessed randomly.
-
Sunday 17th April 2016 16:19 GMT Alan Brown
Re: Bit-Patterned Media
"And they now make drives that still do the work of managing this internally, but with special driver software allowing the layout to be optimized depending on which files are accessed sequentially, and which ones are accessed randomly."
OpenZFS has optimizations built-in for shingled drives.
I still won't use them.
-
-
Monday 11th April 2016 21:13 GMT Anonymous Coward
Put your bean counter shades on...
The are only so many Fabs. It costs big money and a year or two to build or convert a fab for Flash. Flash is a fraction of the market today. Fabs will be built, and Flash will have it's day in the sun (until something better comes along, which looks likely to be 3D xpoint), but it all comes down to production volumes and costs. Seems the spinning rust vendors have consulted with their accountants and have a good idea of where the price points will be two years hence. There will always be a place for spinning rust, if it's cheap enough ...
-
Sunday 17th April 2016 16:32 GMT Alan Brown
Re: Put your bean counter shades on...
"The are only so many Fabs. It costs big money and a year or two to build or convert a fab for Flash."
3D flash can be built on 40-100nm fabs (which already exist in large numbers) with minor mods for the 3D part. Producing flash is likely to be the cheapest way to extend such a fab's lifetime.
3D-Xpoint has been hyped as 1000 times faster than NAND but even El Reg has run the numbers on Intel's published tech specs and realised that it's only 6-10 times faster than current NAND(with larger NAND cells being faster and far more durable than smaller ones, so the endurance claims may end up similarly deflated) with a price point significantly higher than NAND. At this point it will only take off if it can match the price or offer vastly reduced latency.
The imminent death of Flash (NAND or NOR) has been predicted for 30 years and every time the challenger has ended up on the scrap heap as its run out of steam. Bear in mind that you not only have to match current flash when you announce your alternative technology, you have to match or better what's actually in the market when you release it. (The same applies to ferroelectric ram (MRAM) - this has been playing catchup for 35 years, never quite making the grade for large-scale deployment and possibly never will except for specialist applications)
-
Tuesday 12th April 2016 10:01 GMT BinkyTheMagicPaperclip
How much space do consumers need these days?
I'm guessing it's 1TB or less, and that many people could get away with 512GB - which is already hovering at the magic 100 quid price point.
Games are the only storage consumer these days for the average user, surely - everything else is streamed. Pictures don't take up that much room, I wouldn't think everyone is shooting video all the time, even on their phones.
My main system's boot drive is an 840 Pro; 60GB is allocated to Windows 8.1 boot, ~190GB for its apps partition. The main storage is going to be provided by SSD cached RAID10 (4x1TB), I've only just got around to sorting it out, and really haven't needed the space - I'm up to about 180 GB (Windows 8, Windows 8 apps, and data from Windows, Linux, and FreeBSD) with a little stored on an external 2TB drive, the storage from my prior system.
In more than a few years of downloading lots of data, the 2TB drive was only half full. Unless I start ripping all my DVDs and blurays I don't see it filling up. I couldn't justify 4x2TB when buying my local RAID setup (although if I build an external file server, I'll be putting in lots of storage, just because).
-
Tuesday 12th April 2016 14:21 GMT Anonymous Coward
Re: How much space do consumers need these days?
Streaming costs money and data allowance which can be small, plus there's the risk of the source disappearing, so no I don't trust streaming. Plus I record TV shows (in case the shows don't come back) and rip my CD collection (because phones and the like don't have CD drives), so I do amass quite the collection of media files. Plus since I don't want to lose them, I use a mirror in case of catastrophic failure and use parity archiving to control bit rot.
-
Wednesday 13th April 2016 09:15 GMT BinkyTheMagicPaperclip
Re: How much space do consumers need these days?
I'm sure you do, but you're not the average user. Five or more years ago I attended a talk by Team17 at Replay Expo Blackpool, and as an aside they asked who still bought CDs. In a room full of around fifty people, myself and one other person stuck their hand up...
The migration to streamed media has only increased since then.
-
Wednesday 13th April 2016 09:36 GMT Charles 9
Re: How much space do consumers need these days?
You're mixing up "streamed" with "downloaded". An MP3 downloaded from iTunes or whatever isn't streamed but copied to the user's drive, and doesn't have to be transmitted again unless it's from a home server or whatever.
Steam games, for example, are downloaded, not streamed.
-
Thursday 14th April 2016 20:59 GMT BinkyTheMagicPaperclip
Re: How much space do consumers need these days?
No, I'm not mixing them up, as per my original post
'Games are the only storage consumer these days for the average user, surely '
MP3s? Surely most people are using Spotify, Apple Music, or whatever. Same as they're using Netflix, Amazon Instant Video, iPlayer, etc.
-
-
-
-
-
Sunday 17th April 2016 17:36 GMT cloudguy
HDDs always at the edge...
Well, every "trick" is being deployed to increase HDD capacity...helium, SMR and soon HAMR and everything is designed to run right at the edge of failure in order to keep the price as low as possible. The largest capacity HDDs will likely find their best application in object storage environments where their failures can be better managed, but not in desktop or traditional server RAID storage environments where HDD failures at this size would likely be more catastrophic and/or time-consuming to correct. And then there is the rapid increase in SSD capacity. With 15TB SSDs currently becoming available, SSDs have won the capacity race. HDDs still have a cost advantage, but it won't last for much longer. HDD manufactures will likely stop building HDDs sometime between 2020 and 2025. So spin them while you can.
Biting the hand that feeds IT
