Seagate's shingle bathers stalked by HGST's helium HAMR-head sharks A performance-versus-capacity war is brewing in the hard drive industry with Seagate leading on disk-flash hybrids and WD-HGST on helium-filled devices. At a briefing session for tech journos yesterday, Seagate dropped hints of new solid-state hybrid drives (SSHDs) - which combine a non-volatile NAND cache with spinning …
Potentially. That's why most of us peg SMR as a temporary or niche solution. SMR is quite all right for low-write-frequency applications. For example, an external hard drive that uses SMR and is used as a backup device wouldn't see much penalty from the rewriting since writing tasks would be performed in a bulk fashion.
Magnetic media aren't worn out by repeatedly being written. (They can be adversely influenced by the read-write head lying at one particular location most of the time, whether active or not. The mechanism here is interaction between airflow and tiny amounts of contaminants within the sealed HDA. For this reason drive firmware periodically performs a random "elsewhere" seek on an idle disk)
The amount of power used in writing a track is a very tiny fraction of that used to keep the whole mechanical assembly spinning and to move the heads around.
So I wouldn't expect this aspect of SMR to reduce drive life expectancy.
However, there's a major qualitative change with SMR. In a conventional magnetic drive, once your track is written, it stays there on the disk without any maintenance being needed. With SMR, it will be read and rewritten whenever nearby data is updated. If something isn't working right, there is much greater potential for your pre-written data to become scrambled. In this respect, I'd expect SMR drives to "brick" themselves far more easily. More like an SSD (which also performs read-modify-rewrite cycles), less like the magnetic disks we're used to.
I hope that "old technology" drives remain on sale for a considerable overlap period!
So, write head larger than read head, overwrite by reading adjacent area and writing both, domino effect for the width of the band . . .
ouch.
If they can actually get this tech to work with the same data bandwidth and response times that we enjoy on today's disks, in read and write modes, then I will be suitably impressed.
But it is still starting to look like there's a whole barrel of monkeys behind every disk write, and when you introduce more complexity you unavoidably introduce more points of failure.
I won't be totally stupid and say something like "they should reduce the size of the write head" - if they could they most probably would have. But still, it doesn't look too good from the MTBF point of view.
Wait and see, I guess. But I'll be approaching those drives warily.
The problem with head size is that they can't make it smaller while still passing enough current through a coil that small to generate an adequate magnetic field. Too much power in a very small volume = meltdown (and they still haven't found a room-temperature superconductor).
Which is where HAMR comes is. Shine a laser through the magnetic field created by the head, focussed on one track of the several under the head's "bubble" of magnetic field. That heats up the surface of the platter. Choose the magnetic medium right, and the head's magnetic field will flip only the bits in the heated track while leaving the cold tracks on either side unaffected.
To me HAMR feels like good physics and engineering, SMR feels like a kluge. Bits that can only change state when heated should be MORE stable than those stored conventionally.
Some SMR maths. A conventional disk doing small writes may manage 100-up IOPS (seek time ~8ms which dominates, half a rotation of latency 4ms, a bit of optimisation possible by doing out-of-order writes). If they shingle three tracks, the rotational latency of a write goes up to about six revolutions of the disk (three to read it and three to rewrite it after modification). That's about 50ms for a 7200rpm disk. SLOW except for large writes. (Reading will be no slower).
Helium in an HD enclosure interesting twist, however, the big issue is that eventually the helium seeps out. A typical life cycle for a completely sealed glass-enclosed He-Ne laser tube is about 5 years and unless there is a fundamentally new way of sealing the HD enclosure, the He seepage may limit the life of those high-performance drives.
Unless there is something that I haven;'t considered..........
Leakage from a laser tube may be accelerated by the fact that the thing runs hot and the gas is ionised. Is it He2+ that gets into the glass? (He2+ a.k.a. alpha particles, helium nuclei). On the other hand, I'd expect glass to be an intrinsically better gas-container than a multi-part metal HDA.
As has been pointed out two or three times above, monatomic helium can leak through very small holes. If you can build a case to keep helium in, then you can build a case to keep everything else out. Why not evacuate the drives ? There would be even less turbulence and buffeting in a vacuum.
Just a thought - I freely admit I know virtually nothing in practical terms about how you build hard drives.
Chris Cosgrove
Drive heads actually use air "planing" and a miniature form of ground effect to help keep the head floating at the correct height above the platter. No air = no g.e. = no planing. The head would just crash onto the platter with current designs. Read about hard disks at high altitude.
Hard drives actually RELY on air to keep the spacing between platter and head. It's called the Bernoulli Effect. That thin cushion helps to account for inevitable tiny-but-significant imperfections in the platter. At least with helium you can still use the principle (you'd just need to correct for the different gas density).
But as noted, helium is a tricky gas to contain. However, I've read of techniques capable of perfectly capturing hydrogen gas (the only thing smaller than helium: an H2 molecule—hydrogen by itself tends to pair up, as do nitrogen, oxygen, and a few others—has half the atomic weight of a helium atom).
Hydrogen gas has been used in large electrical generators for cooling, etc. Don't know why hydrogen wouldn't be considered unless the spinning rust is susceptible to hydrogen embrittlement. I image the safety hazards of use of hydrogen in consumer-level equipment could also be a consideration.
In a word - heat. Platter and actuator surfaces require some degree of cooling to stop them from overheating. Helium is more thermally conductive than air, thus a better coolant.
Vacuum has terrible thermal properties - there would be no convection heat exchange at all. Vacuuming the disks would reduce the possible density, not increase it.
Didn't think of it that way. I thought the general benefit was a lower viscosity (about 2/3 that of air), so less heat is generated from air resistance. Higher thermal conductivity would just be another plus for helium. Plus, being gaseous, you can still use the Bernoulli Effect.
So what I've learned from this -
1. Seagate is about 4 years behind on HAMR tech as they announced many years ago it was just around the corner.
2. Seagate is having a hard time getting costs down, performance up and a viable consumer market production drive.
3. Even with the new tech Seagate is going to have a hard time getting storage capacity higher any time soon.
4. I better PRAY that SSD storage capacities quadruple in the near future while cost per GB drops significantly.
5. Seagate better have a transition plan to go completely SSD or they'll be out of business within 7 years.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
