Where has the corrections button gone? ....Crabs.
We've seen a lot of innovation in disk drive technology over the past year or so, tech into which millions of dollars of R&D as well as large amounts of time and effort have been poured. But from where we're standing at the beginning of 2014, it may ultimately be totally useless for storage array vendors. I'm thinking of …
Where has the corrections button gone? ....Crabs.
I think it's done to prevent people glossing over the article..
It *is* decypherable, sort of. But someone's been asleep at the helm here.. ;)
To be fare the grammer and the mis use of capitals and not using them for names and stuff make the spellng seem, like a minor problem, dont it ????? !!!!!!!!
Not to, mention; punctuation?
Just one of the perks of online publishing; Crowd-sourced proofreading.
It doesn't work.
It was said that if you put a million monkeys in front of a million typewriters, they'll eventually recreate the works of Shakespeare. The Internet has proved that wrong.
The "almost read only" nature of shingled drives sounds like a good fit for something like NetApp's WAFL - originally designed because they used RAID4, which struggles with writes unless you write a whole stripe at a time. Not at all hard to imagine them offering a shelf of shingled drives as an archival tier, in the same way they like to mix the fast pricey SAS/FC drives with slower, bigger, cheaper SATA ones already. I'm sure lots of data on a typical array doesn't change from one month to the next, but still needs to be accessible in the milliseconds of disk not the minutes of tape - so, shuffle it off in big batches onto a bunch of shingled drives, keeping new or frequently-changed data on regular platters and/or flash.
I doubt a disruption would bring us to the point you couldn't get hold of a replacement HAMR or helium drive for one that had failed - easy enough to keep some spares in stock if you're that worried, and replacing HAMR with helium or vice versa would be a big issue for your array. (Would it? Obviously there's a huge difference between shingled and conventional drives, do HAMR and helium drives differ much in performance?)
How does that notoriously difficult to trap helium stay in there?
Isn't it easier to keep a vacuum than try to keep helium in place?
Every time helium filled drives come up, somebody has to explain, yet again, that the heads have to "fly" on a cushion of gas. It's simply impossible to design a drive using a vacuum as the head fly height is now measured in nanometres on the capacity drives.
It is also perfectly possible to store helium. After all, it's kept in cyclinders under pressure. What you have to do is to create a hermetically sealed environment with no rotating part seals.
I'd also mention that vacuum is much, much thinner than helium, so if you can't make something that can keep helium in, you certainly can't keep a vacuum in. Also as soon as you put anything consumer grade in a vacuum, almost everything outgasses, meaning you end up with no vacuum.
I suppose that they could always fit a turbomolecular pump to every drive, but it might cancel out the capacity density improvements .
Notwithstanding the problem of head levitation, I believe that it is actually easier to maintain a vacuum than to store Helium.
The problem with storing Helium is that it's monatomic i.e. it's normal state is in the form of single atoms, whereas Hydrogen, for example, is molecular and is normally found bound together in pairs. This means that the individual atoms of Helium are smaller than the individual atoms or molecules of any of the other elements with the consequence that Helium will eventually leak through any physical container, regardless of what it's made from and even if it's hermetically sealed.
Maintaining a vacuum, on the other hand, is easier because the size of the molecules present in the air that you're trying to keep out are much larger than the monatomic Helium atoms you're trying to keep in.
I wasn't saying it would be good to use a vacuum in a HDD. Just that sometimes its easier to maintain vacuum than keeping helium trapped.
Helium is a tricky thing.
"heads have to "fly" on a cushion of gas"
We had a drive come in with a damaged head, where the arm was resting comfortably on the disk surface. It wore out a groove through the aluminium platter, in some places all the way through, leaving a semi-circular gap - and yet it was still balanced in place.
Now, if you were ask exactly *HOW* that came to be the customer had the drive powered up the whole time, and didn't use it at all, and didn't actually complain till it "started" making noise...? I have no idea.
"I believe that it is actually easier to maintain a vacuum than to store Helium."
It's the same thing. Since you have to ensure none of the Helium gets out, (and prevent other gasses from getting in), you need to seal the vessle against pressure changes entirely - that's effectively the same as being able to contain a vacuum.
And that's why it took HGST years to make helium-filled drives viable. They are confident that they will be able to keep the helium in for at least 5 years (the warrantee period) on most of the drives. Or they'll be replacing them regularly when they fail.
a Reg article was referring to the Guardian as the Graun just the other day.
El Reg and the Graun have been co-operating recently on security issues. Perhaps there has been a leak in the spelling department.
Or the 'Grauniad' as it's affectionately known.
"They only come from HGST and no other source is available. "
How is that a shortcoming of HGST 6TB drives?
It means they're so innovative no one else has the technology.
What's wrong with that?
Whoever would be using these unique drives would have an advantage over their competitors.
Nothing is wrong with innovation.
Everything is wrong with tying your company to a single supplier.
Single supplier means supply is directly tied to that supplier's ability to deliver. One hiccup on the road, and you have no product.
It also means that contract terms can change at the whim of said supplier. Once the supplier has 100% of its market share in you, in can change pricing basically when it wants. "Due to cost increase on the global market, we are raising our gross price by $10 a unit." Then what do you do ? Shout and stamp your feet ? You have no other supplier. Beg for a delay ? Why would he give one ? Switch to another technology ? That will take some time, during which you have to swallow the change - with the corresponding PR impact and - most probably - the lawsuit the supplier will throw your way for abusive termination of contract.
Production lines have inertia like boats, they take time to retool, and industrial companies are the Titanic, ever on the lookout for whatever kind of iceberg can sink them.
A single supplier is very much an iceberg for any industrial company. So it is not so much having a unique advantage, it is more like having a unique weakness.
I understand not wanting to utilize shingles (because they have technical drawbacks) or the Kinetic drives (because they require that bugaboo, software work), but the argument to avoid helium drives was not very convincing.
Who cares if there is helium inside or not? If your competitors offer the option to put 6 TB drives in their arrays you would be a fool not to sell them for some (probably temporary) single supplier risk reason.
Perhaps the risk comes up in contract negotiations with Hitachi (how to guarantee the risk does not actualize) - but really, it is the same thing as with, say, x86 server processors for the past few years. Intel has had long periods of performance dominance over AMD, but I don't see server vendors refusing to sell any Intel processors that are faster than AMD processors (just in case Intel decides to raise their prices)...
The issue isn't that the drives are helium filled, it's that they are bigger than others and only one manufacturer supplies them. If some OEM depends on 6TB drives for an array, they want to be sure they can supply replacements for their customers (or, if the drives are super popular, have enough to keep the sales channel supplied). They can't do this if HGST is the only one supplying 6TB drives; they are beholden to HGST to manufacture the drives fast enough to keep them supplied. This is the way ALL current drives are handled; all current drives must have a second source (that's why it looks like Seagate and WD are copying one another's drive manufacturing; the OEMs require that they make drives that are virtual copes in terms of capacity, power consumption, and performance, so they can drop them as replacements into RAID arrays without issues).
This is only sensible as a business strategy in light of things like floods in Thailand, earthquakes in Japan, and so forth.
But given that HGST is a wholly-owned subsidiary of Western Digital, it's my bet that we'll see helium-filled drive from WD at some point in the not-too distant future.
You can be assured there are at least three layers of backup for every bit followed by a tape in a fire and water proof safe.
No technology is infallible and no HD unbreakable.
Are these shingled drives a lot cheaper or something? Where is their advantage?
Helium is perfectly able to be stored though, I have a cylinder of the stuff in my garage that I bring out every now and then to entertain the kids, great fun! :-)
Shingled recording is much higher density than regular drives because the tracks can be closer together. So more storage, less space. The tracks being closer together is the reason you can't write a single block; the single block write over-writes adjacent tracks, and the data in them gets disturbed.
Shingled drives are very much better at single write multiple read storage (say, archiving, as mentioned in the article).
Shingled drives are hideously bad in write performance for conventional file systems, which want to re-write blocks in place (which requires reading and then re-writing the entire track they are on). For b-tree based file systems with no-rewrite semantics (which covers some of the newer file systems, like ZFS or MSFT's ReFS), they can potentially be reasonably good performers. And you can bet that the second generation of SMR drives will do something to improve their performance under regular re-write in place file systems (you could potentially virtualize the LBA space within the drive software and re-write blocks to new blocks on the drive, for example; with enough buffer, this wouldn't kill read performance and would improve write performance...)
Sounded shite from the off - because you could clearly see the flaw in attempting to overlap everything - and it of course shows up in the form of abysmal write speeds.
I can't imagine it's a very reliable process either.
Helium filled drives, well, my hat goes off to HGST for trying something new.
Not quite - if you're writing single blocks (or block quantities less than a track or set of tracks), current-gen SMR has bad performance (because of the need to re-write the entire set of tracks associated with that data). If you're writing to a blank drive (or, in the Seagate implementation, writing entire sets of tracks) the performance is as good as current drives since there is no need to read the tracks in before re-writing them, and because the shingled structure of the tracks on the platter is maintained. Of course the problem is that current file systems don't approach writing to drives this way, so they (or the drive firmware acting on their behalf) have to read all of the data on a set of tracks, make a change to the block the file system wants to change, and then re-write the entire set of tracks that contain the data. This really tanks the write performance. File systems that don't re-write in place (copy on write or log-structured file systems) can potentially make better use of SMR if they understand the track set size and honor it (which most probably don't).
Read performance is about what other architecture drives support, once the data is on the drive.
Regular drives have to deal with this issue of writing to adjacent tracks, but since they space the tracks much further apart than a SMR drive does, they have much lower adjacent track effects. Even so, the data written to the drive will decay over time as data is written in adjacent tracks and stray magnetic effects impact the magnetic domains the data is stored in. To avoid these effects, many drives will refresh the data on the surface of the drive by reading it and re-writing it back to the track periodically (this is all done by the drive firmware and is invisible to the user.) This reduces the write adjacent issues and track signal degradation caused by writes on adjacent tracks, and the data stays fresh.
The problem with selling helium drives to upper management:
"But it doesn't FEEL as heavy as the other kind of thing so it CAN'T be as good."
"Helium is too flammable and we shouldn't have that in the computer room."
No, the reason you can't sell helium drives to upper management is that helium is light and so tends to float upwards, like hot air, but helium is even lighter than hot air, so tends to rise faster. Upper management are afraid of competition from this new variety of "Super hot air" and know that it would rapidly replace them at the top of the heap. Helium based drives would obviously result in all the upper management becoming redundant PDQ!
So they can't have that.
The article misses a couple of points. It's predicated on Block-IO and SANs being the future of storage, which is a bit out of touch.
Look at Object Storage and the shingled drive makes sense. In Object stores, data is written sequentially, and if a roll-back to previous versions is needed, any updates create a new block. Shingle drives work really well in this mode, since each written block just tags on the end of the last written block.
Archiving works the same way,and if the runes are read correctly, this is the future for hard drive storage anyway, with Tier 1 being SSD only.
Seagate's new interface makes a lot of sense. It's a lot more efficient than the existing software stack. It's time to revise the stack for all disks.
As for Helium, this isn't a magic substance. All it means is that HGST has figured out how to do a 6TB drive. They've an edge over their competitor, but that's only temporary!
There's a famous (in business geek circles) book called "The Innovator's Dilemma" which is all about introducing new technology to an existing industry that initially the biggest customers in that industry think is crap, only for that technology to get better and better until it becomes the norm, replacing the previous technologies. With the depth of irony that only El Reg can deliver, that book uses the hard drive industry as it's case study! Oh, and it does so with evidence not just opinion.
General purpose arrays write to cache and SSDs, and only occasionally stage out to spinning rust. So the limitation of needing to write in half gigabyte blocks (or whatever the shingle size is) shouldn't be an issue.
Neither is vibration. It would be rather easy for an enterprise array to take measures to isolate drives from vibration, if that's required to get a decent life span from them.
I'd be more concerned whether there's enough of a win in capacity to make it worth the bother. Hard drive density has barely moved in the past few years, and unless shingles or HAMR start moving the needle, I doubt the EMCs of the world, or their customers, are going to be too eager to jump on the bandwagon of a new and not yet well tested technology for a mere 25%-50% capacity increase.
A "mere" 25% - 50% capacity increase???
MERE ? ? ?
ARE YOU KIDDING ? ? ?
Have you read the stats on helium drives? They're something like 45% more energy efficient per unit of storage (per TB). They run cooler, requiring less air conditioning/heat removal, which creates a second tier of energy savings. At 6TB per drive, they require less rack space. And they're NOT going to have SMR's pathetic rewrite speeds.
A drive with those characteristics is a logistical and economic bonanza for the big storage customers.
Shingle is niche. I personally will not knowingly buy a shingled drive. I see it as fundamentally flawed with built-in counter-measures that amount to a handicap. Not so helium.
Assuming that WD/HGST can get 6 to 8 years out of their helium drives -- at which point they will be amusingly small and obsolete -- the helium innovation will be a total winner. It's simple and requires virtually no compromises in drive performance. The helium idea is one that will probably spread down to the consumer level and will most likely be around for the remainder of spinning storage.
I'll also bet that it won't be too long before Seagate has it's own helium drives.
Helium will improve any and every type of spinning drive -- including SMR and HAMR drives. Think about that.
My money is on the "H" technologies: Helium and HAMR.
Biting the hand that feeds IT © 1998–2018