Pluck-filled platter-stuff: Bold disk drive makers fatten up

Analysis Disk drives are getting more platters so they can be fattened up with more data. HDD electric motor maker Nidec thinks 3.5in nearline drives will increase their thickness from today's 1 to 2 inches so they can cram in more data. Seagate's 10TB, helium-filled, Enterprise Capacity drive is 1.028in (26.11mm) thick, the …

[Andy The Hat]

3.5 inch fat-boy

If they reduced the platter diameter slightly, they'd get a pair of spindles diagonally in a 5.14 bay ... double the height and double the spindles with a slight reduction in capacity per stack ... may work.

Or, what if we accepted a reduction in speed as a consequence of greater storage then we could have, thinking out loud here, a "5.25" device in what I will call a "full height" bay. Not sure whether platter stability would impact the density but perhaps 300Tb in a single device with little work?

[Ashley_Pomeroy]

Re: 3.5 inch fat-boy

Wasn't that the idea behind the Quantum Bigfoot? They were slow, but cheap, and they stored a lot of data. I remember a friend of mine had to partition a 10gb model into five 2gb partitions, because those were the days of FAT16.

[Mage]

I have some quite new PCs with "full height" bays (twice CD/DVD Rom). I have full height MFM HDD in the attic too.

Is it 3"?

If so, then why not have a horizontal spindle 6" long with platter diameter between than in a 2.5" and 3.5" drive. Maybe have 15,000 RPM and stripe + parity across sets of four or five platters.

[Natalie Gritpants]

> If so, then why not have a horizontal spindle 6" long with platter diameter between than in a 2.5" and 3.5" drive.

Because the PC will literally fall over when it spins up.

I had a V8 Rover P6 with soggy suspension. It used to rock side to side when you revved it.

[45RPM]

@Natalie Gritpants

Doesn't it depend on how fast the platters are spun up? If it gets spun up slowly then it won't cause the PC to fall over - and, in fact, might make it more stable once it's at full speed owing to the gyroscopic effects of the disk. Of course, it all depends on how they're oriented in the case…

[0laf]

Heee haw

Two sets of platters interleaved wouldn't ake up so much space. You could make them contra-rotate as well to reduce the centrifugal issues.

Sounds like the olden days I was told about once. Of daring technicians wrestling out of balance hard disk drives to the floor as they bounced around the server rooms.

[45RPM]

Re: Heee haw

@Olaf

To make the disks dance, you needed to spin them up and spin them down - and then you could get them to walk (or play tunes). If they were out of balance then they’d shatter when they got spun up - so you can be sure that they were properly balanced in the factory. Of course, I don’t know anything about pre-1987 big iron (so things might have been different in deep pre-history) - and I’ve only ever seen disks shatter once (on a VAX).

[Steven Jones]

Re: Heee haw

Interleaving drives is not an option. HDDs rely on a smooth airflow to fly the heads, and having two interleaved drives will cause horrible turbulence issues.

[Martin an gof]

Cost?

Capacity is all very well, but surely the main force keeping spinning rust alive is the fact that byte-for-byte it is still a heck of a lot cheaper than SSD. Until SSDs can match HDDs for cost there will always be a use-case for HDDs.

M.

[45RPM]

I’ve got a 2inch tall 3.5 inch SCSI drive somewhere. It stores, enormous for its time, a little over 40MB! That’s MB - not GB. The computer I had before, a 4.77MHz 8086 Compaq Deskpro, had a 5MB 5.25 inch hard disk. So these sizes don’t seem desperately alien or surprising to me.

I have to say that reliability of these devices is a concern. The MTBF figures are meaningless - I’m responsible for a system which can ruin disks very quickly. It has 50 of them - and we have about six failures a year. I’ll take a little less capacity in return for bulletproof reliability.

Incidentally, that old 40MB drive, a Connor, still works perfectly.

[phuzz]

5MB on a 5.25" disc? The density on that must be so low you could practically write data with a magnetised needle and a steady hand!

[Andy The Hat]

...

So I presume you believe that the 1Mb IBM hard disc cartridges that were about 16" across were recorded by Fred Flintstone and a chisel?

You have the inate ability to make some people feel sooo old :-)

[imanidiot]

Re: ...

Are you saying they weren't?? But how else could they possibly work?

[Andy The Hat]

Re: ...

hmm ... chisel means lumps, therefore texture, aha! Braille discs! They'd last for ever so would class as 'super-archive class storage'. If the mains power failed you could go fully manual - all you need is one of those unbelievable blind people who can actually read that stuff to shout the output down ye olde style phone ... - ok so the bandwidth drops towards BT Rural Broadband rates but it still works. Now, where did I put my patent-writing pen?

[Anonymous Coward]

Google

Didn't Google ask for this a year or so back?

[Little Mouse]

Huh?

Am I missing something really obvious?

How is a double-height drive that gives less than double capacity better than two industry-standard drives, one atop the other?

[Andy The Hat]

Re: Huh?

If you have two small drives you have just two ordinary small drives, but replace it with a massive one and it's Top Trumps time!

Other than that, I see your point :-)

[Dave@SolidFire]

Re: Huh?

It's cheaper. You get to amortize the shared components (controller board, case, drive motor) over more platters / capacity.

The only thing disk is good for at this point is low $/GB. So every change from now on will 100% focus on that.

[<shakes head>]

Re: Huh?

yes you are missing something the author made an assumption the wring way, that there would be less than double the platters, but as about 1/3 of the drive is circuit board and motors and still only requires the same amount of case.

so 2/3 for case and platters, but now delete 2 layers of case, and you are probably looking at close to 2.5 times the number of platters, not the authors less then 2

[Millwright]

Geometry

Yes, you have to take case/bearing thickness into account when you double the z height - but unless that scales up linearly or worse you'll have relatively more, not less, room inside the higher units.

It's still only breeding faster carthorses to compete with lorries though.

[Richard Simpson]

Re: Geometry

"We can't assume a simple platter count doubling because the casing itself takes up space"

I was confused by this as well. Surely, no matter how tall you make the drive you still have one base and one lid so if you make the drive twice as high you get more than twice the capacity.

If we assume that a 1in high drive has 0.75in of platters and 0.25in of base (including PCB) and lid and that it contains 6 platters then adding another inch gets us another 8 platters. Am I missing something here?

[imanidiot]

Spindle stability

Given the speed the platters spin in a drive I'm not sure the platter spindle/axis length required for such a 2*Z drive would/could be made stiff enough to result in a very robust drive. Platter wobble will become much more of a problem for sure.

[Anonymous Coward]

When drives get this big does that not increase the danger of static head slap errors?

[defiler]

I'm glad that everyone else has pointed out the obvious

If by doubling the drive height you can't double the platters, then just double the number of drives.

If nothing else, it halves your RAID rebuild time compared to a 2" monstrosity.

Besides that, the sweet spot for GB/£ still seems to be 3TB, quite comfortably. If you're worrying about the cost of the controller, perhaps you should store less shit. Just saying...

[Michael Duke]

Re: I'm glad that everyone else has pointed out the obvious

So the sweet spot for 1 drive is probably 3TB.

When installing 120 of those into 12 drive shelves, each shelf needing rack space, SAS cables and cooling in a datacentre environment then density is king. 10RU vs. 20RU when going from 3TB to 6TB pays for the $/GB difference anyway. For the likes of Google/Amazon/Azure/Netflix/Dropbox who run disks in the tens or hundreds of thousands this makes a major difference.

So while for a home PC with 1-2 disks 3TB is the sweet spot, at scale the higher the density the lower the associated costs will be and that is what drives the research.

[defiler]

Re: I'm glad that everyone else has pointed out the obvious

"So while for a home PC with 1-2 disks 3TB is the sweet spot, at scale the higher the density the lower the associated costs will be and that is what drives the research."

True enough. What I will say, though, is that in my experience the hardware costs are dwarfed by software licenses.

However, if you're talking about a drive that's double the physical volume, you can fit half as many into a shelf, rack, aisle, hall or datacentre. I don't see this being a viable solution. It falls into the Quantum Bigfoot dead end.

Luckily, though, it's not my problem to invent a way to fit more into a drive. Just to afford the results...

[Alan Brown]

tall drives - not going to happen

The extra play in the platters will mean the heads won't track properly.

It's the same reason you won't see 5.25" platters. Bigfoot was possible because it spun at 2700rpm and was (relatively) low data density.

[Anonymous Coward]

Found the buyer.

http://www.theregister.co.uk/2016/02/26/google_cloud_disk_white_paper/