Stealthy storage startup wants to fly read-write heads closer to disks Stealthy storage startup L2 Drive's technology could enable higher-capacity disk drives. The company was founded in January 2016 by CEO Karim Kaddeche and COO John Wang. Kaddeche has no direct disk drive storage industry experience, having been a product management director at the Descartes Systems Group. Before that he spent …
Closer?
To my knowledge the head is not even actively positioned, but merely floats on a tiny cushion of air turbulence generated by the spinning disk. Also, a fingerprint on a disk surface would be twice as high in profile as the distance between head and platter. Any of those information tidbits wrong?
With active positioning you'd have to create a head capable of "sailing" undistrubed inside those turbulences, instead of being lifted above them.
Closing this gap even further in a mechanical device seems to be asking for trouble.
I'd heard that HDD heads are about as close to the disk as a Boeing 747 flying a few inches above the runway. And I second the gliding on turbulence bit.
I wish these guys all the luck, but I'm a wee bit skeptical on their chances of success.
That said, whatever they try can just as well bring benefits down the line, so go for it !
Maybe if they run the drive interior under partial vacuum, that will enable active head control to maintain a closer head to surface distance? However, unless they get it into regular mass production, it would be cheaper to buy two ordinary drives that one with special head control.
The current head to platter separation is a minor miracle of aerodynamics. It is self-tuning and works extremely well. Your average drive head gets almost as much wind tunnel hours as a fighter aircraft. It is in fact an aerodynamic problem which is almost as difficult as wing around the supersonic boundary because you do not have real "wind" - the head "flies" in the air flow which is being dragged by the spinning platter. Once it exits the flow (too high) the lack of lift pushes it down, once it gets too close the lift force becomes bigger and pushes it back up.
I have some very serious doubts that a mechanical system under computer control can beat this.
The "drive belt" thing is also odd. Disks used to use linear actuators, driven by loudspeaker-like voice coils. The problem was making a stable linear slide mechanism. Rotary actuators have poorer head to track angular geometry, which is not critical, but much better mechanical stability, which is.
I was going to say much the same thing, but the belts were steel, the actuator was a stepper motor and it was a 20MB drive.
The current head setup more or less _is_ a linear motor. Pivoting on a fixed point rather than in a slide is mechanically simple, far lower friction than a slide and precision than a slide. That patent reads like a leap back to the 1970s.
Disk heads don't so much use aerodynamics as hydrodynamics to achieve a fluid bearing and diskmakers have been saying for a while that the problem isn't ride height, but spot size, which is why they went to shingling and then HAMR (both of which have undesireable secondary characteristics.)
My overwhelming reaction was to wonder if April 1 came around early.
Take a look at the existing HDD technologies that come to the market, many take 10-15 years (or more) to be developed and mature. Does this new technology exist? How long will it take to bring to market and what market is it looking to serve?
Say it takes 5 years to bring to fruition, will it be delivered as a new drive vendor or something Seagate/WD would acquire?
With all these questions, let's think where the flash and Optane markets will be. Flash is outpacing HDD drive capacity increases. QLC will drop costs further. In 5 years, there will be even fewer reasons to use HDDs, so the ROI on the investment of this technology seems very small and un-investable.
I'm a bit surprised by the intended use of a "drive belt".
Given that this will be inside the HDD enclosure, and hence subject to the heat extremes present, I can't think of any device that uses belts, that doesn't need them to be replaced regularly - I'm thinking car engines, certain hi-fi turntables, vacuum cleaners etc etc.
So, given that most HDD's require very precise positioning of the read/write heads, I'm wondering whether any type of drive belt system would work so precisely for the entire expected lifetime of such a HDD.
Invar belts used to be used in HDDs back in the days of stepper motors.
Going back even further, linear motors and rigid slides were common in larger format drives.
Neither of them achieved the necessary levels of accuracy, especially when the compelling requirement is for a smaller head with a smaller spot size and SSDs have already surpassed HDD capacity and beat the snot out of them on price/performance whilst simultaneously closing on raw price/capacity.
They have some freedom now that they can ignore some previous market constraints, due to a lot of the hard drive market going away already or in the near future in favor of SSDs. I think if they target only the enterprise market and VERY high density storage (i.e. minimum 5-10x current hard drives) there's room for success.
SSDs keep coming down in price, but 3D NAND runs out of gas beyond stacks of 96 to 128 layers, and while they may be able to shrink the cells in those stacks the limit they hit around 20nm in 2D NAND still exists. So using current techniques NAND can only get about 5-6x denser than it is today, which puts a lower bound on how cheap it can become per bit. That would make it cheap enough to likely kill off current technology hard drives, but barring advances that can significantly increase NAND density beyond that point, there may be a long term opening for cheaper per bit storage of less performance sensitive data using very high density hard drives.
By targeting only the enterprise market, they can ignore the issues of laptops getting jostled around or PCs under desks accidentally being kicked. Hard drives sitting in a datacenter are very well protected against shocks, so running the head closer to the platter is quite doable in that market. While active control of the head is bound to be more expensive, as well as potentially platters with finer surface smoothness tolerance depending on how close they wanna get, that's not a concern. Going from a base materials cost of <$50 in current hard drives to a higher point (whether that higher point is $100 or $500) in hard drives storing 50 - 100 TB isn't a big deal because they can easily charge a couple grand for a hard drive that large.
"Hard drives sitting in a datacenter are very well protected against shocks"
Not as well as you might think. In dense storage arrays there's a major problem isolating against vibration caused by both head movement in adjacent drives and chassis fans (not to mention loud noises such as people shouting at drives - the average datacentre has a SPL in the region of 90-95dBa)
"platters with finer surface smoothness tolerance"
HDD platters are already amongst the lowest tolerances in production anywhere. The R&D for them costs so much that there's only one maker of the things left. Ditto on drive heads - which is why Seagate and WD both went to shingling more or less simultaneously as the R&D costs could only be justified that way (the differences between them are assembly and secret sauce)
The far more telling thing about HDD technology is that WD and Seagate both disbanded their research units years ago and the the only part left is the arm that's trying to make HAMR work - which is an engineering problem now, not a R&D one. There's no new technology coming down the line.
Trying to justify this on the basis that enterprise will buy it is a non-starter. They're already more expensive than normal drives and SSD has mostly eaten the "low latency" end of the market. If it costs twice as much as existing enterprise drives then buyers will say "fuck it" and move to SSD instead.
AC suggested patent trolling. It's either that or someone's April 1 release got out early.
An earlier Reg article said that read-write heads could not be made any smaller, so closing the air gap is an optimisation that warrants research.
Thinking about this I would ask...
1. Could the read-write head be attached to a flat surface that floats on the air, whilst the head itself is shaped to cut through the air nearer the disc surface? (Or would such a head cause/receive damage when parked?)
2. Could the need for gas be eliminated by using magetism to position the read-write arm (obviously the magnets wouldn't be near the read-write head)?
3. Did researchers ever try extending the read-write arm beyond the disc to run in a groove that maintains it's elevation over the disc surface? If mechanical wear of the grooves is a problem, perhaps a combination of (2) and (3) would work?
4. Could magnetic interference between two adjacent heads on the same arm be used to read-write data at varying depths within the disc?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
