Seagate hits 1 terabit per square inch, 60TB hard drives on their way
extremetech.com
extremetech.com
Main content is: The maximum capacity of today’s 3.5-inch hard drives is 3 terabytes (TB), at about 620 gigabits per square inch, while 2.5-inch drives top out at 750 gigabytes (GB), or roughly 500 gigabits per square inch. The first generation of HAMR drives, at just over 1 terabit per square inch, will likely more than double these capacities – to 6TB for 3.5-inch drives and 2TB for 2.5-inch models. The technology offers a scale of capacity growth never before possible, with a theoretical areal density limit ranging from 5 to 10 terabits per square inch – 30TB to 60TB for 3.5-inch drives and 10TB to 20TB for 2.5-inch drives.
Seagate's press release makes it sound like there're no other HDD manufacturers in the world besides them.
Another way to look at it is to see how many bare drives are being sold at those specs (1TB 2.5") vs. external drives. The bare drives won't sell as well if they are a non-standard height (at least I assume they wouldn't).
Many laptops will also accept the thick 12mm drives you're referring to.
I'd say it's a big deal that Seagate claims they will develop 60TB HDs with your usual exponential growth curve (bearing in mind we're starting at 2-6TB.) "On the way" doesn't mean tomorrow.
Well the problem there is that it's a theoretical upper bound, which is another way of saying "we have not yet found a specific reason that 3.5 inch 60TB drives should be impossible with this technology". That is a far cry from "on their way".
So now, a top of the line Xeon is faster, has 10 multi-threaded cores, includes AES acceleration, and uses 20% less power than it's predecessor 3 years ago... thanks to that base technology.
If you're running a business that generates/consumes alot of storage, this announcement means that you need to start thinking about how to deal with a petabyte of storage capacity with performance characteristics similar to what you have today. (for 1/30th capacity)
Today you can fit about 96T into 4U if you squeeze a little. That's 48 spindles then, sharing the load.
With 60T drives that turns into 2.8P on the same 48 spindles.
At today's maximum SAS bus-speeds (roughly 10 GBit/s) it would take 33 days just to write such an array full (assuming all components can sustain that throughput).
And let's not get started on IOPS...
Anyway, it will be a while before 60T drives hit the shelves. Until then at least the bus-speed issues will probably be sorted out.
For sure, if you get a degausser or other very strong field you will cause damage. But there are two strong magnets inside the case near the platters, so I'm not sure what kind of magnetic field is available in common domestic or office settings that might be harmful.
The dropoff depends on how far apart the ends of the poles are. (i.e. from a distance the two poles blend together and average to zero), it also depends on the orientation - are the poles parallel or perpendicular?
I may be conservative, though. I mean, I also still instinctively gauge program size by how many floppies it would take up . . .
Same with photography. Each raw photo from my 5D II is 25MB. The megapixel race isn't over.
New and more interesting error correcting bits are going to be needed here. I could imagine a device with 60TB of raw capacity providing 20TB of data you could read reliably though.
WD lists their 2.5TB drive as having "Non-recoverable read errors per bits read" as "<1 in 10^14" and the drive has about 2 x 10^13 bits, so while the chance of an error in one whole disk read is non-trivial, it's not 100%.
Mirroring will still work fine as long as the error is detected. The chance of read errors in the exact same sector of two disks should still be small. Silent errors were always a problem for mirroring.
At what point do we as consumers start considering drives to be defective? Is the underlying ECC suffering because of the need to hit greater storage numbers?
I don't see the rate at which we use space slowing down any time soon, but we have physical magnetic platter limits to worry about going forward.
We will always have room for more space!
Not sure I would trust the wear and tear of an alloy that is constantly heated and cooled.
60TB of write-rarely cheap storage for backups would be very handy though.
Today, yes, but this isn't going to give us 60TB drives today, or even tomorrow. It's going to give us 6TB drives. It might give us 60TB drives at some point down the line, and it would be unwise to make any strong predictions about workloads at indeterminate points in the future.
Assuming a 100-year lifespan, if every second of your life were recorded at 5mbps, you would need just 30 such drives.
Yeah, I'm pretty confident about my prediction as to most workloads.
I find it kind of difficult discussing 60 TB since it will be soo faar in the future. By then we will at least have 4k and probably a bitrate in ~100 MBit/s (since the cap is at 50 MBit/s today that is aiming low but there will probably be a limit where you say that it is enough (then again, 3D imagery could potentially demand much more)), a small drive (same density but ~2") would hold about 2 weeks of footage which IMO sounds reasonable for a camera in an age where you don't have to compromise that much.
It will put some constraints on usage patterns, but I don't feel that 60 TB will be that much more data than 4 TB is today. Sure, for text and audio you basically will have an infinite amount of space (as a consumer) but that is the case even today. 60 TB is only 15 times more than we have to day in a single drive, and it will take many many years to get there. And I can barely fit my grandfathers VHS camera tapes on a 4 TB drive today, and since I want some redundancy 4 TB isn't enough for even that.
Blu-Ray bitrates are bloated for various reasons, like older encoders, film transfer, and a lack of anything better to do with the space. You don't need 20mbps with modern H.264 High encoders, it's just a waste. Anything past 10 is at best questionable
> By then we will at least have 4k and probably a bitrate in ~100 MBit/s
Even if 50mbps were something that we would ever need at 1080p (it's not), 4k would not justify a doubling in compressed size, there is no reason to have 100mbps compressed video.
4k is also way beyond reasonable for home display. Just 3840x2160 on a 60" display would be something like 23 arc seconds at 3 meters, you physically cannot see that.
> 60 TB is only 15 times more than we have to day in a single drive
4TB is already orders of magnitude more than most workloads will ever need. I was using video to demonstrate the absurdity of thinking typical workloads would actively use 60TB, because it's not a typical workload, it's a worst-case scenario.
We would need those high-resolution displays if we ever want to be able to show off pictures decently, and we have (finally) just begun to experiment with high-resolution displays. 4k displays are already being demoed.
My point was that 60 TB vs. 4 TB, given the time difference, isn't such a big deal. Biggest hurdle is to get a stable COW-filesystem. But as today SSDs will rule the workstation (and pretty much everywhere you value performance) and SSDs have much bigger technical challenges to get that big.
Regular spinning harddrives will thus only be used where you have a untypical workload requiring lots of data.
But most people won't want the hassle of re-encoding the data they receive. If it takes up twice as much storage, so what? A lot of people will buy that 60TB drive instead of a 30TB one.
> Even if 50mbps were something that we would ever need at 1080p (it's not), 4k would not justify a doubling in compressed size, there is no reason to have 100mbps compressed video.
So consider 4K + twice the frames to handle 3D + doubled framerate. That's 16 times the amount of uncompressed pixels. And I'm not convinced that's the upper bound of the amount of raw data we'll see.
It's not that long ago you'd have been called crazy if you suggested we'd need anything beyond 1080p, and a lot of people were questioning whether HD would ever get any traction.
> 4k is also way beyond reasonable for home display. Just 3840x2160 on a 60" display would be something like 23 arc seconds at 3 meters, you physically cannot see that.
I know people with projectors. 60" is by no means the upper bound on size people will want. 100"-150" maybe.
Even my phone records video at a higher bitrate than most DVD's use, and my camera even more.
Nearly every grandparent I've run across has a >100GB hard drive (if not >500GB). They're all using roughly 20GB of that drive, after a couple years.
With only very rare exceptions, people don't choose what goes into their computer. They buy what looks good for a price that they're willing to pay, for the simple reason that buying a computer is full of more buzz-words, acronyms, and useless numbers than any other purchase I've ever made. It's insane, and it gets people to buy more than they need, so the sellers won't stop doing it.
But what will a computer look like in 10 years?
What I've been noting, from the whole mobile explosion, is that people really don't care if their computer only has 64GB of storage. Of all the complaints that people have about tablets and phones replacing 'real' computing, disk space rarely enters the discussion.
So even if tablets and smartphones don't notably replace traditional PCs, their effect on consumer perception of the necessity of large disks can't be ignored.
Sure, it's easy to market 'more gigabytes than your old machine' to a regular person. But so is "faster boot" and "less waiting".
That's why laptop makers can't seem to ditch spinning disks fast enough.
And what motivation does a PC builder have, to continue bundling increasingly-capacious hard drives with little markup? Why not bundle increasingly-faster and increasingly-cheaper SSDs and save the spinning drives for higher-margin positions as add-ons and external storage?
I'd be incredibly surprised if the market for spinning disks doesn't shrink over the next 10 years. They'll certainly remain, even for consumers. But there's no good argument that they'll still be standard.
I totally agree that spinning disks are on the way out, except for high-storage purposes. And then only until flash storage meets / exceeds their density (solid state has a habit of doing this). And I totally agree that the HD size means nothing to most people buying computers. But it does drive sales over smaller numbers, so we'll keep seeing them go up as long as that's true. Maybe the tipping point is now, but I'm not putting any money on that.
As for 'boots faster / less waiting', it isn't something that's generally quantifiable because it comes with a wide range of caveats. Even if it were (I haven't seen any, but I haven't PC shopped for a while), most people I know are well aware that computers slow down over time. They may not know why, but they have seen it happen to every computer they've ever owned - "It wasn't this slow when I bought it" is a common complaint. On top of that, the vast majority of people I know simply don't shut down their computer until an update forces them to. Computers resume from sleep very quickly - my laptop is awake and responding by the time I can get my hands on the keyboard to punch in my password. Cutting that time in half gains nothing, it has reached the 'fast enough' point that it's not an incentive.
I will wait for empirical data myself.
As for reliability - I don't think it's going to be that much of an issue - I've had CD-RWs that I used as portable storage and wrote/erased/read dozens of times (with scratches on them from constant use). A fixed platter in a clean housing should last quite a while...
http://www.youtube.com/watch?v=-xPvD0Z9kz8 (Get perpendicular! animation by Hitachi)
EDIT: Here's the original video direct from their website (higher quality): https://www1.hitachigst.com/hdd/research/recording_head/pr/P...
How soon until Dell, HPs (etc) ship more desktops with SSDs than with mechanical drives? The average consumer doesn't understand the difference and doesn't want to pay extra (even though it's probably the best performance/$ upgrade possible).
On the flip side, I wouldn't be surprised if we saw significant adoption for servers that can live with the space limit (or that can design around it). Server-based applications are often I/O bound, established solutions are expensive, and the people responsible know and care.
For example, my laptop has a 250GB HDD and I have a few (read: too many) external hard drives that add up to about 10TB.
I would love to replace my laptop drive with a SSD and the external drives with one or two of these new 6TB drives.
SSDs are great, but that doesn't some people won't also need big HDDs.
Write techniques may or may not need to be adjusted (preheat a track? a sector? realtime per bit-domain?) but should eventually converge to the same speed-up as reading.
http://en.wikipedia.org/wiki/Magneto-optical_drive
The MO drives use a laser to heat the recording substrate to the Curie temperature during the recording path, too.
(Ok, sorry, couldn't resist)
Without mention of the power required for the laser it's only speculation.