Seagate Adopts Helium For a 10TB HDD (computerworld.com) 175
Lucas123 writes: Seagate has finally adopted helium as an inert gas in its data center drives and has used it to produce a 10TB HDD for cloud-based data centers. Seagate had relied on its shingled magnetic recording technology for high-capacity drives right up until its last 8TB HDD, even after WD has used helium in several iterations of its hermetically sealed, 3.5-in HDDs. The lighter-than-air helium reduces friction on platters and allows more to be used. In Seagate's new HDD, it crammed seven platters 14 heads, a 25% increase in disk density over its 8TB drive.
Oh yeah! (Score:2)
Can't wait to see the failure rate on this thing. How do they even get a hermetic seal?
Re: (Score:2)
Draw a couple of snakes on a sealion?
No, just bits of synthetic rubber and screw down on them really tight. If the pressure difference isn't huge between external and internal it's not very hard.
Re: (Score:2, Troll)
If you knew anything whatsoever about helium permeation, you would know what a preposterous statement you made.
It's the difference in PARTIAL pressure that matters. There is almost certainly one atmosphere of helium on the inside vs zero on the outside. About the same as a rubber circus balloon. You know, the kind that lose all their lift in a day or t
Diffusion over distance (Score:2)
Re: (Score:2)
Can't wait to see the failure rate on this thing. How do they even get a hermetic seal?
Well we've been doing this with car tires for ~15 years now, nearly 20 on luxury model cars. Despite that both the rubber, the rubber-steel/aluminum bonding point, and the valve all leak. The solution was to over pressurize the tires in the early years, these days they simply slap some sealing compound on it. It's generally good for the lifetime of the tire as long as it's done properly. That's between 3-8 years.
I'm guessing it'll be similar in this case, though they'll likely have an easier time of it
Re: (Score:2)
If we are talking about car tires rhen they don't need the lightest inert gas. Argon would do the job and its not as rare (on earth) as He
the real problem.. "high" bit set (Score:2)
and the data comes out funny.
Re:Oh yeah! (Score:5, Interesting)
There's significant and there's long enough (Score:4, Interesting)
Re:Oh yeah! (Score:5, Informative)
> These drives will leak.
While technically correct, the rate of static-pressure helium leakage through HGST HelioSeal appears to be measured in decades. They up-rated their enterprise SAS drives from 1.4 million hours MTBF to 2.5 million hours MTBF because hermetically-sealing drives and using helium improves various operating parameters, prolonging life in several ways.
My results in production and the lab bear this out over the past two years: helium drives appear to have substantially lower failure rates than air-filled drives. While nobody has owned a commercial helium drive for a decade yet, the internal helium sensors on the disk farms that I've looked at show no degradation or leakage so far: SMART 22 shows 100.
I'll be watching Seagate's results here with great interest and optimism that their results parallel those of HGST.
Disclaimer: I'm an Oracle employee; my opinions do not necessarily reflect those of Oracle or its affiliates.
Re: (Score:2)
an arbitrary number mapped to a measured value you will never see. 100 != 100%
How do they handle pressure changes? (Score:3)
If you have had hands-on experience with these things: do you have any idea how they handle pressure differences? Hard disk casings typically have flat sides, which are not ideal for handling pressure. In a pressurized airplane, you can have 0.3 bar pressire drop, which will exert about 30 kg of force on the walls. It's worse if they are shipped as unpressurozed cargo. So, the walls must be thicker than with a conventiomal HD, which conflicts with the goal of increasing the number of platters.
Re: (Score:2, Funny)
... the internal helium sensors on the disk farms that I've looked at show no degradation or leakage so far: SMART 22 shows 100. ...
Hooray.
Those sensors were manufactured by Volkswagen....
Re: (Score:2)
Disclaimer: I'm an Oracle employee; my opinions do not necessarily reflect those of Oracle or its affiliates.
And all credibility of your post was lost with this sentence. You might want to keep that secret, NO ONE thinks Oracle is impressive anymore, we discovered that you don't have to live in the seventies or pay for features that everyone else just does by default.
Re: (Score:2)
Re: (Score:2)
> MTBF means nothing, because of the realy stupid way it's calculated...
I'm not disagreeing with you, but isn't MTBF -- when combined with warranty and AFR -- an overall expression of a hard drive manufacturer's engineering confidence in the product?
Backblaze among others have noted that it's really difficult to estimate failure rates because the sample sizes are too small at present (see https://www.backblaze.com/blog... [backblaze.com] ). So it's fair to say the jury is still out as to whether the decrease in operati
Re: (Score:2)
No, manufacturers "confidence" is pure marketing. And your mention of Seagate, who bought the Maxtor production lines and used them to produce "Seagate" drives with seriously fake MTBF numbers, makes this all the more funny.
Car manufacturers will drive a test vehicle real distances, not "estimated", because if their products failed at the rate of hard drives, they'd be out of business. And offering a refurbished drive in place of a brand new DOA (Seagate's done this to me 4 times) is also crap. Imagine doi
Re: (Score:2)
MTBF for drives is like the contrast ratios they like to put on LCD panels. It's a 100% pure bullshit made up marketing number. This has been proven by studies done by companies like Google who statistically significant numbers of drives, and their real world MTBF is nowhere close to the manufacturer's numbers.
Re: (Score:2)
Dimensional mismatch. XT^-1 != T
Correction on why Helium "leaks" (Score:3)
Re: (Score:2)
I think He is smaller and more diffusion prone in many ways than hydrogen.
I believe that the difference is hydrogen is soluable in many metals. So it dissolves in the metal as monatomic hydrogen then diffuses through. The reason for this is not because H2 is smaller than He, but because H is sort of a metal and is soluable in many metals.
For non metals, I think that doesn't apply. I might be wrong though.
Re: (Score:2)
Since I've never heard of such a thing, please provide a link.
Note: given that hydrogen, even when diatomic, is smaller than Helium, it doesn't need to be "soluble" in metal to diffuse better.
Re: (Score:3)
Hydrogen has a higher diffusivity than helium. Seems to be about 25% higher.[PDF]
That seems to be about diffusion of gaseous things within H2, not diffusion of H2 through solids.
Since I've never heard of such a thing, please provide a link.
https://en.wikipedia.org/wiki/... [wikipedia.org]
http://energy.gov/sites/prod/f... [energy.gov]
http://www.google.co.uk/search... [google.co.uk]
Note: given that hydrogen, even when diatomic, is smaller than Helium,
Really? The atomic radius of Helium is 31pm http://periodictable.com/Prope... [periodictable.com] whereas the bond length
Re: (Score:3)
Sealing helium in ANYTHING for a significant amount of time is pretty much impossible. Helium is a monatomic gas. These drives will leak.
Leak where? Yes the atoms diffuse through the metal but then what happens to the drive? They aren't pressurized, they are just filled with helium. In order for them to leak other particles need to diffuse into the drive to replace the missing atoms, it's not going to leak itself into a vacuum, just like helium balloons never completely and totally deflate, they just get to the point where the rubber is no longer exerting pressure on the gas.
Re: (Score:2)
"In order for them to leak other particles need to diffuse into the drive to replace the missing atoms, it's not going to leak itself into a vacuum,"
Actually, you will draw a vacuum, provided that you wait long enough. In balloons, the process slows down not because the pressure difference decreases, but because the latex rubber layer gets thicker as the balloon shrinks.
Re: (Score:3)
"In order for them to leak other particles need to diffuse into the drive to replace the missing atoms, it's not going to leak itself into a vacuum,"
Actually, you will draw a vacuum, provided that you wait long enough. In balloons, the process slows down not because the pressure difference decreases, but because the latex rubber layer gets thicker as the balloon shrinks.
Citation please. The only reference I've found including in physics textbooks is that helium can diffuse through metals. I haven't found anything that says you can hermetically seal helium in a metal container and then come after a while to find it has drawn a vacuum.
Re:Oh yeah! (Score:5, Informative)
Let me start with an appeal to authority: I actually get paid to do calculations on gas diffusion and pumping of hydrogen.
Depending on the materials of the walls of your helium-containing vessels, drawing a vacuum can take rather long. The point is that diffusive transport is driven by differences in partial pressures (or concentration if the gas is dissolved in a solid). The partial pressure of helium in the atmosphere is about 0.5 Pa; if you have a vessel with a porous wall with 100 kPa of helium (atmospheric pressure) on the inside, then helium will diffuse towards the volume with the lower partial pressure until both sides have the same partial pressure (i.e., 0.5 Pa). The same process will happen in the opposite direction for other gases (nitrogen, oxygen), but at a much slower speed. So at t=0, you have 100 kPa He (pure). After 1 year, you have (for example) 50 kPa He and 0.01 Pa nitrogen. After ten million years, you have 0.5 Pa He and close to 100 kPa nitrogen.
Just imagine that you have a box with a small hole and lot of fruit flies on the inside. Place this box next to a stable where there are lots of big flies. The fruit flies will gradually disappear from the box, but not because they are pushing each other or because the fat flies (that don't fit through the hole) are pushing them out.
Here are the basics of diffusion: https://en.wikipedia.org/wiki/... [wikipedia.org] .
For helium, diffusion speed is proportional to the difference in partial pressures on either side of the wall. For hydrogen, it's more complicated because the hydrogen molecules first need to dissociate before they can permeate through metals; it turns out that the speed of diffusion is driven by the difference in square roots of the partial pressure of hydrogen on either side.
Re: (Score:2)
Thanks for the explanation. It actually clicked when you mentioned partial pressures. I was thinking in absolutes.
Re: (Score:2)
Re: (Score:3)
In an inflated balloon, you have 1.05 bar of He on the inside and 0.000005 bar of He on the outside.
When the balloon is almost deflated, you have 1.00 bar of He on the inside and 0.000005 bar on the outside. So that would make the diffusion speed decrease by a whopping 5 percent.
Re: (Score:2)
And why you compare the pressure of He inside, to the partial pressure of He outside is beyond me. It is a gas, under pressure, pushing against a membrane...with the gas gradually going through the membrane. You could i
Re: (Score:3)
Because the pressure of He inside is equal to the partial pressure of He inside.
Unless the pores in the balloon have diameters that are much larger than the mean free path length (about 50 nm at atmospheric pressure), the absolute pressure doesn't matter for the leak rate, only the partial pressure. I doubt that balloon rubber has pores that large, but it could be. Even in the likely case that the pores are sm
Re: (Score:2)
Re: (Score:2)
Built in Obsolescence. Exactly what they intended =p
Re: (Score:2)
Re:Oh yeah! (Score:4, Funny)
404 not found. I'm guessing that means it leaks.
Re: (Score:2)
No, no, that means the leak was properly prevented. You'd know it leaked if the information was actually available outside the company. :-)
Re:Oh yeah! (Score:5, Interesting)
Wow why wouldn't the skilled engineers at Seagate and Western Digital think of stuff like that?
They should employ you on the spot!
Re: (Score:3)
Wow why wouldn't the skilled engineers at Seagate and Western Digital think of stuff like that?
It's because the employees are clowns. I mean literally. They confused the disks with party balloons and started squiirting helium in so they'd float around the room at parties.
Obviously they need a slashdotter to come along and set them right.
Re: (Score:2)
i realize this was probably intended as a joke but...
"Hermetically sealed" is a reference to the hermetic heretics who, among other things, practiced what most people would call alchemy and one part of that involved sealing something air-tight. Long story short, if you wanted something to have an air-tight seal you wanted it hermetically sealed and the phrase outlasted the heretics by a long, long time.
(As an aside, they didn't consider themselves to be heretics but felt that their beliefs were entirely con
Re: (Score:2)
Mean time between failure is not representative of real life experience. Ever. Also, the below-atmosphere internal pressure compared to stp and lower density of the gas makes head crashes a bigger problem than before. That's why you don't run hard disks in a vacuum - they wouldn't work.
It seems likely that running the head closer to the platter is one of the methods they are using to get the data density they want.
Disks we use now would not work at all in a vacuum. The heads "fly" over the platter using some gas-fluid physics to do it properly.
Those drives are going to fail a lot, helium is indeed hard as hell to keep contained. It leaks through a lot of materials that stop many other gasses.
Re: (Score:2)
Did they do away with the breather hole and paper filter?
Re: (Score:2, Insightful)
Of course. The drives have to be sealed as tight as a helium tank.
A side benefit is that you can submerse helium drives in an inert coolant.
Actually, wouldn't doing so greatly reduce the leak rate?
Sure helium is monatomic, but it would be surrounded by presumably some kind of mineral oil, or similar material. That could be used to keep the temperature pretty constant, so no great pressure delta would presumably be formed..
Re: (Score:2)
Until the advent of helium hard drives no hard drive had *EVER* been hermetically sealed. There is always a small hole which allows air to be exchanged with the surroundings so the pressure in the drive equalizes with the outside. There is a small filter to make sure any air entering the drive is kept clean.
I suggest you dismantle a hard drive and see for yourself.
Re: (Score:2)
You don't even have to disassemble the drive. The hole [howtogeek.com] is marked on the top of the drive to make sure you don't block it. It's necessary for proper operation.
Re: (Score:2)
That link tells you absolutely nothing other than they have a patent on the seal. I wouldn't even think of buying one of these drives.
Re: (Score:2)
Why would it require going back to 5" drives? SSDs have higher data density than HDDs, they just have lower data density per dollar.
Re:Oh yeah! (Score:5, Informative)
The amount of data you can fit in a normal drive would be a good measure. Some googling says the largest 3.5" HDD is 10TB; the largest 3.5" SSD is 16TB.
Re: (Score:3)
The people making spinning platter drives are just grasping at straws at this point. It won't be more than a decade before SDDs completely take over. You can already get a 1 TB SSD for about $350. A 1 TB HDD costs about $50. That's a ratio of 7:1. SSD prices have been falling a lot over the past year or two, while HDD prices have remained pretty much constant. There's very little reason for most average people to even be using HDDs at the moment apart from people who want to store giant media collection
Re: (Score:2)
You can already get a 1 TB SSD for about $350. A 1 TB HDD costs about $50. That's a ratio of 7:1.
That's comparison favors the SSD, and it's still 7:1. Look at price per gb for various models/sizes. If you ignore performance, HDD's are still a very large way ahead.
There's very little reason for most average people to even be using HDDs at the moment apart from people who want to store giant media collections.
... or anything that takes up much space. Backups, photos, video's, movies, music, virtual machines, etc etc - all the stuff people actually use desktops for.
IMO, it would be more accurate to say that there is a very good reason for the majority of people to be using at least one SSD, as they do bring significant performance gains, and you are
Re: (Score:3)
They already have non-consumer 10TB SSD's available in the 2.5 inch form factor. I don't even want to know what the price tag is on those things, but they're supposed to be out there.
Re: (Score:2)
Wasn't aware of that, didn't see them when I checked really quick.
Them being 'really out there' would be a valid reason.
Hydrogen next? (Score:5, Interesting)
We need to assume that hydrogen will be the next element used for cooling? Or is it the end of spinning disc era?
Hydrogen is used, believe it or not, for generator cooling at power plants. Here is the quick link: https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
Hydrogen is not an inert element..
Re:Hydrogen next? (Score:4, Insightful)
Re: (Score:2)
Buy hydrogen also likes to dissolve in many metals. At that point it becomes worse for containment than the larger helium. Still though, these things are well established and there's ways of keeping hydrogen contained too.
Re:Hydrogen next? (Score:5, Informative)
In hard drives, the fill gas is used to lift the heads, not for cooling. The idea is that the thin film between the head and platter forms at a shorter distance in helium, so everything can be made smaller and closer together. As another poster pointed out, at room temperature/pressure, helium is monatomic while hydrogen forms H2 molecules, which are larger than the helium atoms.
Re:Hydrogen next? (Score:5, Informative)
What matters for the hydrodynamics (drag forces, lift forces on the head) is not directly the size of the molecule, but the molecular mass (related to densitiy of the gas) and the dynamic viscosity (related to both molecular mass and molecular size). The size of the molecule or atom is in any case vanishingly small compared to the distance between the head and the platter. The dynamic viscosities of a few gases at room temperature are: helium is 19 micro-Pa s, air 18 uPa s, and hydrogen 9 uPa s. The molecular masses (proportional to density) are 4, 29, and 2, respectively; this is where helium wins, but hydrogen is better both in molecular mass and viscosity.
The real reason for not using hydrogen gas is that hydrogen (H2) is reactive; at surfaces, it tends to split up into hydrogen atoms (H), which can then diffuse through metals and polymer seals. In the best case, it will leak out within months/years. In the worst case, it will change the crystal lattice and cause material failure. In particular, rare-earth magnents tend to crumble if exposed to hydrogen gas; that's something you really don't want inside a hard disk casing.
Re: (Score:2)
The spinning disk era is coming to a close, and I welcome it! The issue is that while storage capacity has, for decades, increased almost exponentially, the actual performance of the HDD has remained virtually flat. A typical HDD spinning at about 7200 RPM can store 4 TB of data or more, but can only serve about 150 seek operations per second. Physics, she is a bitch, you know? So while you might be able to store 500 million files, it takes a month to copy them.
Everywhere I look, Enterprise or "performant"
Re: (Score:2)
Everywhere I look, Enterprise or "performant" storage has moved to SSDs.
You may want to define performant. Yes SSDs are used a lot in high activity enterprise scenarios. But they currently have close to zero marketshare in the SAN / NAS department where space and density is more important than speed of access due to being orders of magnitude more expensive.
I'll welcome the demise of the classical harddisk when SSDs are comparable in price, and not a moment sooner.
Careful (Score:4, Funny)
Last time I bought a helium hard drive, it floated away and I never saw it again.
Re:Careful (Score:5, Funny)
Re:Careful (Score:5, Funny)
So what you are telling us is that you are storing your data in the cloud?
Re: (Score:2)
So what you are telling us is that you are storing your data in the cloud?
Not only that but it does actually rain in his DC
Re: (Score:2)
Re: (Score:2)
Well, I guess you could try spinning it up and yanking the power repeatedly until the emergency parking shears off the heads on the park ramp. Then repeatedly seek the drive to the outermost track and hope that the remains of the head manage to puncture the outside of the enclosure. Full disclosure: you'll probably have to seek the drive by manually applying voltage to the stepper motor, because the drive probably won't even show up on the SATA bus if it can't read track 0.
Worth a try, anyway. :-D
Re: (Score:2)
Just set all the bits to 1. Ones are heavier than zeroes. It should be enough to counteract the helium.
Re: (Score:2)
Isn't Helium running out? (Score:2)
Re:Isn't Helium running out? (Score:4, Interesting)
How many 10TB hard drives do think Seagate will ship?
How many children will have at least one balloon at their birthday party this year?
Re: (Score:3)
actually alarmist nonsense that we're running out of helium. The truth is that most helium is just vented off the top of natural gas deposits even now. We have a helium wasting problem.
Re: (Score:2)
Yes there is source we can extract in significant volume though currently at 1000 times the cost. The same atmosphere where we vented it.
Re: (Score:2)
Balloons count for 19% of Helium consumption each year. A standard tank of Helium fills about 50 balloons for a party. According to the PDF here [hgst.com] a single tank fills 10,000 drives. Seems minimal.
Re: (Score:2)
I heard the same thing a few years ago and briefly got all hot under the collar about party balloons ... then I realised that helium is being continuously replenished as a by-product of radioactive decay (alpha decay) [wikipedia.org], plenty of which is going on within the earth's mantle and crust.
Now I can sleep soundly again, and dream of partial pressure shells / helium shell airships sailing majestically through our skies.
Re: (Score:2)
Re: (Score:2)
Finite? Depleting?
The Sun creates 1.2 trillion tons of helium a second. There's plenty of it. The only hard part is getting it. Which is why we need a crash program to either deep-drill or strip-mine the Sun for the valuable helium it's hiding from us.
Re: (Score:2)
This brings up an interesting side topic: now that we're exploiting more natural gas than ever before, are we finding any more helium?
SMR was a DOA idea (Score:4, Insightful)
Re: (Score:2)
Re: (Score:2)
I've been looking for inexpensive 8tb drives for nvr's what I have read is that the SMR drives can't handle the load of 24/7 video recording at 4MBps.
Plus I don't think you could run a raid 6 or even a raid 1 array on seagate drives that large and be able to replace them faster than they failed.
I certainly wouldn't want to trust a single drive with 8tb of data.
Re: (Score:2)
I've been looking for inexpensive 8tb drives for nvr's what I have read is that the SMR drives can't handle the load of 24/7 video recording at 4MBps.
They can just fine as long as they're not fragmented. Continuous copying is a non issue. Now video editing on the other hand.... the nature of SMR is that it needs to rewrite adjacent tracks so random reads and random writes performance is poor.
Plus I don't think you could run a raid 6 or even a raid 1 array on seagate drives that large and be able to replace them faster than they failed.
Seagate is plagued with lemons. From what I can tell the 8TB archival drive isn't one. But I'll wait for more data to come in on that. I'm not actually certain about running these things in RAID1 or RAID6 anyway. SMR needs complicated controllers on the drive to work
Re: (Score:2)
The major decrease in performance depends entirely on the workload. For archival purposes and long continuous unfragmented writes they perform pretty solidly compared to normal harddisks.
For random reads and writes I question why you would want a classic harddisk anyway. SMR drives worked well for storage of large files and would work even better if the OS was aware of them and used them appropriately (think of them as the very first SSDs where the OS didn't align sectors or issue TRIM commands, there's unu
Helium (Score:2)
Cloud-based data centers? (Score:2)
Why only cloud-based data centers? Are they not reliable enough for actual data centers?
Re: (Score:2)
Re: (Score:2)
Why not use a vacuum? (Score:2)
Re:Why not use a vacuum? (Score:4, Informative)
Re: (Score:2)
Vacating the air to replace with helium does not mean drawing a vacuum. You can achieve the same thing with a helium purge.
The helium will leak out, and drives will fai(#)&a (Score:5, Informative)
Worked for an measurement instrument company building instruments that had to work in helium atmosphere. We tried for a long time to seal the helium out. Even to the point of filling the entire inside with glass filled epoxy to prevent intrusions of helium. In the end we gave up, and did a redesign to work in helium. solid metal seals will work, but pretty much any other seal will not.
Re:The helium will leak out, and drives will fai(# (Score:2)
What pressure were you working with? Helium will happily diffuse through metals but it is not going to draw a vacuum on itself. Yes the drives will leak gently, but it's not like a pressure vessel where which is literally constantly releasing a small stream of helium through the shell. HGST hermetically seal their drives after filling them with helium and claim the helium won't leak out within the expected lifetime of the drive (~10 years).
Spishak Mach 10TB drives (Score:4, Interesting)
I don't like the idea of cramming platters to increase density because it throws a wrench in useful scaling relationship between density and I/O rate. I don't want a disk requiring days to sync up or otherwise doubles time needed to read out a given percentage of the disk. This is what archival media is for.
Would much rather see R&D efforts focused on increasing density and therefore I/O performance of individual platters otherwise for my purposes better off simply buying more and scaling out disks.
If helium increases reliability over long term use then great.. if it lasts only as long as the warrantee period I'm not interested.
Hoping against hope something not resembling vaporware will come out of RRAM efforts like crossbar in the next year or two.
Umm... (Score:2)
...can you even hermetically seal helium? It will leak over time, slowly, no matter what.
Re: (Score:2)
the more you say "cloud" the stupider it sounds (Score:2)
"Cloud-based" datacenters? So, what, these drives wouldn't be any good for the datacenter in the basement with no internet connection?
Well, then (Score:2)
MOAR PLATES! (Score:2)
Re: (Score:2)
Yes, the air cushion is an issue. I am sure it can be solved, though. Remember that the head assembly of the cheapest optical drive maintains micron accuracy for both tracking and focus distance while the disk wobbles with each rotation (it's never really centered or rotates in-plane). The head is mounted on voice coils and uses active feedback loops.
If you use vacuum and passive magnetic levitation bearings the energy to keep the disk rotating drops to virtually zero. You can have cold storage that is read
Re: (Score:2)
Remember that the head assembly of the cheapest optical drive maintains micron accuracy
With modern hard drive, according to Wikipedia, the flight height of the head is at 3 nanometers. I actually didn't believe it at first and had to check a few other sources to convince myself that there weren't a few zeroes missing. There is even research for subnanometer flight height.
To get down to such ridiculously small values, the flight height is controlled using thermal expansion based techniques, using a small heating element inside the head.
Re: (Score:2)
One bar is not a tremendous pressure.