"That's not a typo. 10,000GB on spinning platters in helium"
Ten terabytes? I can remember when you could run a tactical nuclear air base on 80 Mb.
HGST today launched a full frontal assault on the market with a shedload of disk and flash announcements – including a 10TB hard drive. The background scene set is the notion that the rate of data growth is doubling every two years and the spread of internet connectivity to non-IT devices (aka Internet of Things) will fuel …
No.
He didn't mention locations, dates, units, specific weapons or other equipment, procedures, or anything else identifiable, he didn't even mention which service branch it was, and it could be any of them but the Coast Guard. Even then, if its not classified and/or covered by a Non-Disclosure Agreement then there's no breach.
This post has been deleted by its author
This reminds me of that recent SSD review where I got downvoted for pointing out that bad sequential write performance undermines the point of buying an SSD.
It doesn't matter if the SSD is big and cheap if it can't keep up with my spinning rust.
That SSD from the review wasn't any faster than my current spinning rust.
The rest of us will be taking advantage of what is available while you're over there getting blue trying to holdout.
"sequential write performance"
What are you doing that uses sequential write? Ever business app I can think of - let alone stacks upon stacks of virtualised servers - is "I/O blender"-style randomness. About the only thing I can think of where sequential write makes a difference is single-user video editing, if you are saving to the local system.
But even then, only if you're using RAID 1 or JBOD. Any other RAID levels will benefit more from something that doesn't suck at random I/O more than they will from sequential...
I don't know, man. I look at storage prices of rust drives today and even I can't complain about price. Even SSDs are at an acceptable level. We've reached a point at the consumer and SMB level where the cost of the media is not the barrier.
Is it has a tendency to seep through *anything* - even solid steel / aluminum..
A balloon lasts a day or two - a foil one a week or two, while an 80 litre dewar flask of liquid helium will noticeably ebb away over the course of a few months. Granted this is primarily via the oil film seal around the stainless steel ball valve once it has been opened and closed again (rather than the metal itself, though that does happen), but I suspect the slightest knock/vibration could disturb the seal on a helium-filled hard drive, enough for the helium to gradually percolate through the seams of the case and up to the heavens - and your data soon follows it!
my friends who make Deuterium bullets (for fusion) use recycling helium cryogenics, so there might be some sealing that is more effective at room temperature (vs supercooled).
Then there is graphene of course....!
Anyone knows what is the helium life in these drives? It would give us a clue on how it is sealed... from a web site on helium leak testing, 10^-10 mbar.l/s is 1 cm^3 in 300 years...
P.
Maybe this will make Seagate pull their finger out of their back sides and actually give us the vapor wear that is HAMR and drop this shingling rubbish like a stone.
Seagate + Shingling + Wiring Data = Dead in under 6 months, ive had to stop supplying them as a result... and Switch over to WD blues and blacks.
Good on you HGST, at least someone pushing things forward.
I can tell you first hand since seagate introduced shingling ive seen the average life of drives when used in raid arrays go from 3 - 6 years to less than 1 year.
I must have gone through 30 + drives before i got sick and tired of pulling another seagate DM model.
Swapped to using WD blues and blacks where people could afford them and ive not had a single bad drive in over 3 years....
In machines where data wasn't saved in ext4 and/or also didn't change a lot they made it past the one year mark but still then had a fair few go in under 2 years and some strange faults that magically went away after low level formatting the drives (which appeared to be shingled drives more sensitive to temperature changes).
Shingling is a stupid technology, change one bit and then your re-writing until then next shingle gap, massively increasing drive wear, yeah it means you can produce bigger drives, but at the cost of reducing the reliability.
Its interesting that no sooner had seagate introduced shingling they very shortly followed with the reduction of warranty to 1 year...
Thats just my experience...
"I can tell you first hand since seagate introduced shingling ive seen the average life of drives when used in raid arrays go from 3 - 6 years to less than 1 year."
That is surprising since Seagate released its first shingled drive (SMR) in 2014, and it is not clear whether those are actually shipping to end-users right now or not (it appears that most SMRs have been shipped to a cloud provider). Your observed failure rate had nothing to do with SMR, and you probably have no SMR disks. Your other observations are probably similarly accurate.
Considering the ever-decreasing supply, and its vital uses (MRI scanners being a notable example), this seems wasteful for that last iota of performance vs a more common gas or putting the effort into increasing SSD capacity and performance (probably need helium to manufacture SSD's anyway, but maybe more TB's per helium unit?).
For a "status report":
http://www.decodedscience.com/helium-shortage-situation-update-one-year-later/42314
We've had this discussion before. It uses very little helium per drive, and may be recoverable (they spent 7 years developing a seal that would keep the helium in). It's not an iota of performance that is gained, it's more like 20% less power consumption. I'm not sure why you think SSDs need helium for manufacture; the point of the helium HDD is to reduce the friction caused by the spinning parts found in hard DISK drives.
Maybe someone needs to come up with a 1 children's balloon = X helium-filled HDDs figure so we don't get this alarmist talk about helium on every helium drive story. Helium-filled drives are not going to affect the "crisis" (which is more of a free market issue).
my brain insists on showing me a rack full of little mylar party balloons . . . each straining at it's connector trying to float upwards . . . then some fool opens the rack door and all the data is stuck to the ceiling a few seconds later.
on a slightly more serious note . . . wouldn't reducing the internal pressure of the drive enclosure help reduce the rate of helium loss? . . . you might have to add little "wings" to the R/W heads to get them to fly at the correct height in a lower density gas but that seems easy enough and would reduce friction from the platters even more.
and what about recharging/purging ports on the case to replace lost helium? . . . pressure change could be monitored and if a drive is losing or gaining pressure or friction losses begin to rise or fall the S.M.A.R.T. software, or equivalent, could just alert the need for maintenance . . . it seems that this would eliminate any issues related to the helium causing end of life problems.
[oh well . . . it's probably more complicated than that so i'll just return to contemplating folks chasing their data all over the ceiling]
Reducing the pressure will reduce the diffusion rate, but that will happen anyway if you start with helium at atmospheric pressure. Helium will be lost at a much faster rate than air is gained, so the internal pressure will decrease. What's needed here is a new head technology that can work in a vacuum from the start. Then we could forget about helium. In the long run we'll only use solid state storage. In the meantime this reminds me of the '60s when electronic controllers took a long time to overtake mechnical controllers. In the face of competition, clockwork made vast strides in reliability, complexity and cost. IIRC, washing machines didn't start to get electronic (microprocessor) controllers until about 1990! Industrial controllers went electronic before then.
The drowned factory of WD left the industry with some breathing space of not having to bring out new products immediately. So they could invest more in research... which now shows fruits.
Other than that, we're seeing incremental progress, not particularly exciting, but it'll be done.
Just had two (!!!) 3TB fail within a week with total fail on spin up.
My external 3TB is the same design and this is getting backed up as we speak.
One thing that seems interesting about HD drives like this is the use of a piezo actuator for the head motion means that they could be attacked using ultrasonics on a resonant frequency if this can be fed back the head could indeed be interfered with or worse, hit the platter during long write cycles.
An important thing to note is that some drives use the same buggy FW so by using RAIDed drives made by different manufacturers with near identical sizes then the odds of a simultaneous failure are much reduced ignoring power spikes (UPS) or other unforeseen events.