Muted HAMR blow from Seagate: damp squib drive coming in 2016 Seagate R&D bigwig Jan-Ulrich Thiele says the first Seagate prototype drives built with heat-assisted magnetic recording (HAMR) will arrive in late 2016 and have just 4TB capacity, according to Heise.de. This is not quite right as what Seagate says is: "Seagate anticipates shipping a small number of HAMR qualification units to …
No amount of increasing of bit density is going to resolve the fundamental problem with HDDs, and that's IOPs and, of course, latency. Indeed the higher the areal bit density, the worse (in relative terms) the problem gets. I suppose if HDDs are just to end up as a semi-archival repository for very rarely accessed data then price/capacity is the only advantage left, so it makes sense. It might well be in five years time HDDs are reduced to a niche product.
Helium filled or not, the fundamental issues with HDDs remain as they are dictated by geometry and materials and engineering can only make marginal improvements in ameliorating the issues. Adding caching helps to a certain level (whether device, array, system or application based) helps, but at a certain point it hits the law of diminishing returns. The future looks bleak for the makers of spinning stuff of all sorts of storage, whether it's HDDs, CDs, DVDs or anything related.
You don't need tens of thousands of IOPS for a media drive or cold storage tier. Price per TB is still unmatched. A 15 terabyte SSD is nice but it will cost thousands of dollars for now.
If HAMR can knock HDD prices to below $10/TB and boost capacity up to 100 TB, there is still some life for the hard drive.
The costs of the chips will go down. How long that takes depends on demand and their ability to create the chips.
With non voltile resistive ram (NVRe Ram) on the fringe ... we can see the end of the spinning rust.
I remember starting out with my Ohio Scientific 3A and 2 double sided drives that stored 256K per 8" floppy back in the late 70's. So sorry, I will not miss the spinning rust, but welcome the advancement of new tech. smaller, faster and using less energy.
You are making the mistake of assuming HDDs are trying to compete on IOPS. They're not (at least since SSDs went mainstream). They're large data drives. Your new 30GB game (most games are large indexed archives), multi-TB video collection, TBs of photos, etc. You don't get 8TB of spinning rust to put your OS and apps on.
Hard drives are still excellent with the one thing they're good at: large sequential writes and reads. HAMR will even improve the other thing they're good at: offline (or nearly so) data storage, as NAND requires data refresh cycles (similar to RAM, but with a larger timeframe between refreshing), which means data on an SSD that's tossed in a drawer will (most likely) not be readable in 3 years.
There is no inevitable leakage. At least not at any rate that will matter during the lifetime of a drive. The engineers who design these drives will be aware of the issues. These aren't party balloons.
If you doubt helium can be contained for very long periods, consider how the stuff is actually extracted. It's from separation from natural gas reserves where it accumulates over millions of years from the radioactive decay of uranium and thorium. The same conditions that trap natural gas also trap the helium. If helium can be trapped over millions of years (under pressure), there won't be any problem in containing a small amount in an aluminium casting. I suspect it's the gasket which is the only real problem.
"Helium diffuses in aluminum"
Yes and that thesis is about when it's pressurised. Hard drives aren't, so diffusion rates are extremely low.
You don't have to prevent helium diffusing out indefinitely, just make sure that it takes at least 7-8 years to do so to any noticeable extent.
Helium doesn't just diffuse through aluminium either; when studying I learnt that photomultiplier tubes (essentially a glass vacuum-tube with a metal suitably susceptible to the photo-electric effect inside, and some clever, if simple, amplification) shouldn't be stored where helium is present. The helium diffuses through the glass and ruins the vacuum, if given enough time.
Glass is a solid, for any meaningful definition. It's just not a crystalline solid. It's a class of solid matter commonly referred to as amorphous, or often just as a glass.
Another class of solid is made of long tangled chain molecules, such as rubber or some plastics. There may be further categories.
"Helium-filled drives from HGST come with 5 year warranties."
Not really viable for long-term archival storage then. :)
Does it have to be Helium? Wouldn't Neon do? It's about as unreactive - unlike that dangerously volatile Xenon character!
Perhaps we could make the drive casings out of "transparent aluminum(sic)" and have dual-purpose devices: data-storage and illumination?
Besides the "helium leakage" issue, I don't suppose enough people realize (and the industry isn't going around telling them, of course) that flash memory has this annoying problem of just getting amnesia over time.
One of the scariest aspects of SSDs and flash memory in general, to me, is the fact that data can just start randomly disappearing while they sit on the shelf, with no voltage applied.
Someone at Seagate who also is a member of the JEDEC standards-body wrote a highly publicized paper on this which was published last May, but the issue of offline flash/SSD data retention has been known-about for years.
One of the scariest aspects of SSDs and flash memory in general, to me, is the fact that data can just start randomly disappearing while they sit on the shelf, with no voltage applied.
You think that can't happen with a hard disk drive? Or with tape? I've seen both suffer bit-rot in storage. With a hard drive it can be utterly catastrophic (the drive won't spin up again, possibly because its bearings have siezed).
FWIW I've not yet experienced a DVD-R stored in a dark cupboard going bad on me. OTOH it's probably just a matter of time, and that's not a practical way of storing Terabytes in any case.
The only "safe" long-term storage is a continuously running and monitored storage system with multiple redundancies, data-scrubbing, and prompt replacement of any failed or failing disk drive followed by data regeneration. Which costs.
If HAMR drives need a larger gap between platters to accommodate a laser in the read-write head assembly, how can Helium reduce the gap if the fundamental problem is the head size? To solve the head size problem, either the lasers must be shrunk in size, or they must be moved outside the platters and the energy transferred to the head via some kind of a heat resistant optical fibre.
On the issue of hard drives v high capacity SSDs, the primary issue is cost, not data density. Multi terabyte SSDs cost thousands of pounds because they comprise many, perhaps hundreds of layers of silicon, each of which have to individually etched. Unless the cost of the chip fabrication process can be reduced through automation, then large capacity SSDs will remain expensive. Also current SSD technology is approaching a limit, so don't expect capacities to continue increasing for long, until a new memory technology replaces it.
If HAMR drives need a larger gap between platters to accommodate a laser in the read-write head assembly...
That would be such a major and fundamental design problem that the proposed technology would never have gotten off the drawing board.
Well, probably. You can cite the Austin Allegro, the DC10, and Windows 8 as exceptions that prove the rule.
I'll give Seagate the benefit of the doubt for now. It's their company on the line ....
What about using 3.6mm lasers and optics located on the side of the platter?
The idea here is that the beam can be set to fixed focus and with some clever design and a simple on-chip waveguide be turned to hit the platter where needed.
Plus the advantage with this method is that if the laser diodes start to weaken they can be replaced without affecting anything else in a cleanroom.
Bonus points for including laser powering of the head amplifiers so more than one head can be on a given head arm for increased IOPS.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
