KentuckyFC writes "The density of magnetic memory depends on the size of the magnetic domains used to store bits. The current state-of-the-art uses cobalt-based grains some 8nm across, each containing about 50,000 atoms. Materials scientists think they can shrink the grains to 15,000 atoms but any smaller than that and the crystal structure of the grains is lost. That's a problem because the cobalt has to be arranged in a hexagonal close packing structure to ensure the stability of its magnetic field. Otherwise the field can spontaneously reverse and the data is lost. Now a group of German physicists say they can trick a pair of cobalt atoms into thinking they are in a hexagonal close packing structure by bonding them to a hexagonal carbon ring such as graphene or benzene. That's handy because the magnetic field associated with cobalt dimers is calculated to be far more stable than the field in a cobalt grain. And graphene and benzene rings are only 0.5 nm across, a size that could allow an increase in memory density of three orders of magnitude."
Makes me think of the time I downloaded a very large app disguised as a.wav file - after three hours of downloading, Windows kindly told me it wasn't a valid wav - poof, gone!!
Next time try Save As instead of Run and that won't happen. (BTW the file was most likely still in your TEMP folder, just named something unrecognizable. But recovery would have been possible most likely.)
It's already a challenge to fill a 60GB MP3 player with MP3s. I have 9TB of disks on the network at home, and it's less than half full, even with all of our CDs and DVDs ripped onto the server - and of the 9TB, we use 6TB as double backup of the 3TB primary storage.
What's a person to do when disk capacities increase by another 3 orders of magnitude?
You are thinking wrong. Instead of thinking of disk capacities increasing by 3 orders of magnitude, think of disks as shrinking nice and small (1 1/2"), using a lot less power and generating less heat yet being faster and storing twice the data of today's drives. Netbooks with the storage capacity of a large desktop of today.
Let me guess. They're going to stick this stuff to a platter and spin it past some sort of electromagnet. I want terabyte USB thumb drives, not yet another mechanical storage device.
I would worry too much. Tape seems to be on the way out because it can't keep up with the density requirements (Data silos anyone?)
Some places now just mirror to other hard drives. -Some are smart and take those HDs off line and treat them like tapes. -Some are idiots and leave them online only to find them corrupt like the main disks (I read that somewhere...)
Anyhow, it seems we are going to mimic Star Trek and just not bother to have backups for computer systems...
No problems here, we have two of each generation of tape drive (prod+dr) and when we upgrade we retire them into storage instead of throwing them into the landfill. If our drives were inoperable or we couldn't get them hooked up easily there are companies out there that specialize in retrieval from tape. Beyond that the failure rate for LTO in my experience is vanishingly low, we put over 100 tapes a month through our libraries and I think we've had two failed tapes in the last 3 years and one of those was dropped on its edge so completely understandable. All our tapes go through verification on a different drive than wrote them and we do test restores both at prod and DR. My experience with DLT was almost as good. If you use anything cheaper than DLT then you aren't really using tape meant to be reliable IMHO. The farthest back we have been asked to go was 15 years for documents related to property taxes which can apparently be refiled for up to 20 years in some jurisdiction, no problem recovering the DLT tapes (well, there were filesystem and format problems, but nothing related to the tapes and even those were fairly easily overcome).
Check out the BER on any modern tape and compare it to the BER on any hard disk. For example - Current model Seagate Barracuda ES drives - 1x10^-15. Current model HP LTO drives 1x10^-17.
That's two orders of magnitude better. Furthermore, consumer grade disks which are significantly cheaper (and thus competitive with tape) tend to be an additional order of magnitude worse.
Especially problematic when the tape drive dies. How many tape drives do you have in stock?
The thing is, I can get new tape drives, as many as necessary, in fact. With hard drives, you pretty much get the one shot (or it becomes impractically expensive).
Anyhow, it seems we are going to mimic Star Trek and just not bother to have backups for computer systems...
Or just put it up online and somebody somewhere will have a cached version of it within a few days. All you need to do is link to your webserver from your blog.
Anyhow, it seems we are going to mimic Star Trek and just not bother to have backups for computer systems...
The Enterprise keeps backups in a protected archive in the computer core. In Contagion [memory-alpha.org], La Forge restores the corrupted memory caused by an Iconian probe by shutting down the computers, wiping the memory, and restoring systems from the protected archives.
I want terabyte USB thumb drives, not yet another mechanical storage device.
That day may come sooner than I'd thought. It looks like they even have 256Gb thumb drives now, last time I checked the largest was only 32Gb (which is now the sweet spot in $/GB). I'd still like to have all the data I own on 1 disk with another disk or two as backup. If mechanical gets me there sooner so be it. Currently mechanical is 1/20 of the cost of USB flash, comparing lowest $/GB media. 10Tb HDDs should be here in 2013, acc
German physicists say they can trick a pair of cobalt atoms into thinking they are in a hexagonal close packing structure by bonding them to a hexagonal carbon ring such as graphene or benzene
I have a friend who was tricked into thinking he was a hexagonal close packing structure after spending a bit too much time around benzene.
say they can trick a pair of cobalt atoms into thinking they are in a hexagonal close packing structure by bonding them to a hexagonal carbon ring such as graphene or benzene.
...
the cobalt has to be arranged in a hexagonal close packing structure to ensure the stability of its magnetic field. Otherwise the field can spontaneously reverse and the data is lost.
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with
2-methyl-1,3,5-trinitrobenzene
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with 2-methyl-1,3,5-trinitrobenzene
Meh. That'll only be a problem for the overclockers with liquid nitrogen cooling. The rest of us will just end up with a pile of cobalt and a bunch of hexamethyl chickenwire.
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with 2-methyl-1,3,5-trinitrobenzene
Actually, the explosive yield is greater if you omit the methyl group. Trinitrobenzene out-booms trinitrotoluene, but is harder to handle due to its lower stability.
the cobalt has to be arranged in a hexagonal close packing structure to ensure the stability of its magnetic field. Otherwise the field can spontaneously reverse and the data is lost.
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with 2-methyl-1,3,5-trinitrobenzene
Yeah, and not to mention that cobalt atoms can be very nasty [wikipedia.org] if they decide to isotope themselves. Will the EPA to allow them in PCs?
Well this is all fine and dandy for storage space, but what about performance? Drives are getting bigger all the time, but you're still stuck with spindles that rotate at the same speeds as the ones from last year, or the year before. I can't see anyone wanting to replace their speedy many-spindle database disk farm with a single 320TB disk that still spins at 10Krpm and only delivers ~125 IOPS. Performance is going to suck big time. All the top TPS benchmark results for example are achieved using 100
As data density increases, so does the rate at which it can be read. Assuming two orders of magnitude increase (100x) and individual bits staying roughly the same shape, the linear density increases by a single order of magnitude. (10x bits per track, 10x tracks). The drive will be able to read at 10x the speed.
At 3 orders of magnitude, you can expect a read speed improvement of roughly 3000%. (sqrt(1000) ~31.6)
Ok sequential IO is going to improve some as more data will pass under the head compared to todays disks. But Random IO isn't going to feel the same benefit because that's influenced more by Rotational Delay (fixed by spin speed) and the time it takes for the head to shift between tracks: Disk Seek. So your figures are going to be wildly off in real life.
Worst case seek time is limited by rotational latency and the speed you can slam the head from the inside of the platter to the outside and get it under control. Luckily with increased density we are able to shrink the platters needed to achieve a given amount of storage thus decreasing the latter metric, the first should be shrinkable with decreased mass but the best I've seen is 20k rpm drives which never really took off. It's a very real limitation in today's enterprise, SSD's largely solve the problem b
Back when I was in college one of the 'cool' old Comp Sci professors had a tale he liked to share with his classes on the first day. I had him in a couple of classes, so I heard it over and over again. His presentation made it an amusing story if you could get over the fact that he smelt as if he lived in an ashtray.
It seems that back in the mainframe days, the standard way of increasing storage size on your hard drives was to make a bigger platter. Seems rather simple, right? The storage size grows exponentially with its radius. So adding an inch each time can lead to some fairly nice results, and with some platters topping out at 24 inches, that's some space.
Except....
One day, the university ordered the 'latest' hard drive for one of their mainframes. I'm sure it was a behemoth, it probably held around 50 meg. The vendor came by and installed it, and everything seemed fine till a few months later when the drive seemed to start failing, at about 30% capacity, writes stopped working and anything written to seemed to have been corrupted. They were puzzled, but this is why such things service contracts. The vendor came out, replaced the drive, and everyone went on with life.
Till it happened again, at about the same capacity. Another replacement was made and vendor was quite red-faced and explained that they seemed to have run into a batch of dud drives. All was forgiven and life went on.
Till, it happened the third time. At this point, it was starting to embarrass everyone: The vendor, the people who ordered the hard drive in the first place, etc. So this time, instead of just allowing the vendor to take the drive back, the dean of the department demanded they diagnose the issue there on the spot.
Now, this wasn't the age of the sealed drive cases, certainly drives were still kept 'clean' but we weren't to the point yet where a single grain of dust could wipe out megabytes of info (heck, even the 24 inch platters needed to be in arrays of 50+ just dream of hitting 100 meg) so cracking open the drive wasn't that big of a deal.
So the vendor's tech, hoping to appease a clearly angry customer in the day and age when parts cost tens of thousands of dollars, popped open the drive.
Want to guess what they found?
Larger disks do indeed result in more surface area, but they also result in a higher centrifugal force on the edges. An increased force which the vendor apparently hadn't accounted for. Once the disks began to spin up, the glue holding the magnetic dust to the platter gave way, resulting in the platters being stripped clean after a certain radial length from the center. The disks themselves were fine up to that point, the dust was plastered to the case itself and when the platters came up to speed any dust that had fallen back onto them was once again flung up against the case.
The reason why the disks didn't seem to fail till they reached a certain capacity was simply because they weren't being used in a RAM fashion but were being written to in a sequential manner. The outer portions of the platters were only being hit once the inner portions were written to.
Perhaps the reason spindle speeds haven't gone up lately could be part of the same issue. Or perhaps I'm simply indulging in a bit of pointless nostalgia as I wait for this report I'm running to finish. Who knows...?
The storage size grows exponentially with its radius.
At a fixed data density (and a fixed number of platters), storage is proportional to the area of the platter, which is proportional to the square of the radius.
Storage size grows quadratically with its radius, not exponentially.
Not that it's necessarily the case, but I'd imagine there is some minimal spacing dictated by the strength of the magnetic fields in use. The smaller structures could allow for smaller spacing than would be allowable for their contemporaries.
Before I can get excited, I need to know when this is proven experimentally. The FTA refers to a calculation. There are lots of possible things that are achieved with a calculation, but translating it in practice is a totally different matter. BTW, I am an experimentalist nanoscientist (working on graphene, actually), part of my daily job is to prove that computational results can be achieved in reality.
Sure, but if it were to somehow leave the device, that would mean there are cells that can no longer function. I can't imagine this could make it in the market if any appreciable amount benzene could be "lost" after manufacturing.
Though, if my math is correct, you only need about 1 picogram of Benzene for a Gigabyte worth of cells. Not worth considering even if it wasn't locked away inside a silicon and ceramic package...
1000x Denser ??? ... 50,000 atoms... to 15,000 atoms... 8nm across... only 0.5nm across... could allow an increase in memory density of three orders of magnitude three orders of magnitude? what kind of math is this??? The only question now is whether this team's calculations hold true in the real world. I would like to see that calculation!
This may be the breakthrough, though, that allows for the type of density that would be required for a human-analog type AI to be a reality.(currently it would take a small building to approximate a typical human brain)
Yes, that's damn good. Three orders of magnitude is very roughly a full decade's worth of progress in the hard drive world. Whoever did the graphene work has really earned their pay.
More room, yay! (Score:4, Funny)
Is there a difference? (Score:4, Funny)
Sweet, more room for p0rn. I mean, more room to store my philosophical musings about the world we live in...
And the difference is what again?
Parent
Re:Is there a difference? (Score:4, Funny)
There are more seeders for porn.
Parent
Re: (Score:2)
Sweet, more room for p0rn. I mean, more room to store my philosophical musings about the world we live in...
And the difference is what again?
One is for masturbation, the other is mental masturbation.
Who controls magnetism... (Score:4, Interesting)
-- I'm just not sure he knew exactly how that would come out to be true!
Re: (Score:2)
-- Magneto
Re: (Score:2)
Re: (Score:3, Funny)
i just rename good files as porn image files.
That way my data is hid by obscurity. And since it is porn it is freely shared thus backing up my data like a real man. by having the world mirror it.
Re: (Score:2, Informative)
Re: (Score:3, Informative)
How to fill up the storage? (Score:2)
What's a person to do when disk capacities increase by another 3 orders of magnitude?
Re:How to fill up the storage? (Score:5, Insightful)
Storage requirement is going up, relentlessly:
VCD = 700Mb
DVD = 4.7Gb
Dual-layer Blu-ray = 50Gb (potentially 100Gb; 4 layer @ 25Gb per layer.
And don't forget that Ch-erman scientists never sleep:-)
Parent
Re: (Score:3, Informative)
You are thinking wrong. Instead of thinking of disk capacities increasing by 3 orders of magnitude, think of disks as shrinking nice and small (1 1/2"), using a lot less power and generating less heat yet being faster and storing twice the data of today's drives. Netbooks with the storage capacity of a large desktop of today.
Not again! (Score:5, Insightful)
Re: (Score:3, Interesting)
I would worry too much.
Tape seems to be on the way out because it can't keep up with the density requirements (Data silos anyone?)
Some places now just mirror to other hard drives.
-Some are smart and take those HDs off line and treat them like tapes.
-Some are idiots and leave them online only to find them corrupt like the main disks (I read that somewhere...)
Anyhow, it seems we are going to mimic Star Trek and just not bother to have backups for computer systems...
Re:Not again! (Score:5, Interesting)
Tape is still very much "in" if you're talking about long term storage.
Parent
Re: (Score:2)
Indeed, I know of a government institution that just invested in a 7 loader tape library (read room).
Tapes are not dead, and anyone who ignores their worth ends up paying the price.
Re:Not again! (Score:5, Informative)
Parent
Re: (Score:3, Informative)
Long term - my ass! Reliable - bah! Cheap? no.
Hard-drives are surprisingly superior.
Check out the BER on any modern tape and compare it to the BER on any hard disk.
For example -
Current model Seagate Barracuda ES drives - 1x10^-15.
Current model HP LTO drives 1x10^-17.
That's two orders of magnitude better. Furthermore, consumer grade disks which are significantly cheaper (and thus competitive with tape) tend to be an additional order of magnitude worse.
Re: (Score:2)
The thing is, I can get new tape drives, as many as necessary, in fact. With hard drives, you pretty much get the one shot (or it becomes impractically expensive).
Re: (Score:2)
Anyhow, it seems we are going to mimic Star Trek and just not bother to have backups for computer systems...
Or just put it up online and somebody somewhere will have a cached version of it within a few days. All you need to do is link to your webserver from your blog.
Re: (Score:3, Informative)
The Enterprise keeps backups in a protected archive in the computer core. In Contagion [memory-alpha.org], La Forge restores the corrupted memory caused by an Iconian probe by shutting down the computers, wiping the memory, and restoring systems from the protected archives.
Re: (Score:2)
Re: (Score:2)
That day may come sooner than I'd thought. It looks like they even have 256Gb thumb drives now, last time I checked the largest was only 32Gb (which is now the sweet spot in $/GB). I'd still like to have all the data I own on 1 disk with another disk or two as backup. If mechanical gets me there sooner so be it. Currently mechanical is 1/20 of the cost of USB flash, comparing lowest $/GB media. 10Tb HDDs should be here in 2013, acc
What a coincidence! (Score:5, Funny)
German physicists say they can trick a pair of cobalt atoms into thinking they are in a hexagonal close packing structure by bonding them to a hexagonal carbon ring such as graphene or benzene
I have a friend who was tricked into thinking he was a hexagonal close packing structure after spending a bit too much time around benzene.
Lets show 'em what a walkman can do! (Score:2, Funny)
More room but---- (Score:4, Funny)
say they can trick a pair of cobalt atoms into thinking they are in a hexagonal close packing structure by bonding them to a hexagonal carbon ring such as graphene or benzene.
the cobalt has to be arranged in a hexagonal close packing structure to ensure the stability of its magnetic field. Otherwise the field can spontaneously reverse and the data is lost.
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with 2-methyl-1,3,5-trinitrobenzene
Re: (Score:2)
Meh. That'll only be a problem for the overclockers with liquid nitrogen cooling. The rest of us will just end up with a pile of cobalt and a bunch of hexamethyl chickenwire.
Re:More room but---- (Score:5, Informative)
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with 2-methyl-1,3,5-trinitrobenzene
Actually, the explosive yield is greater if you omit the methyl group. Trinitrobenzene out-booms trinitrotoluene, but is harder to handle due to its lower stability.
Parent
Re: (Score:2)
the cobalt has to be arranged in a hexagonal close packing structure to ensure the stability of its magnetic field. Otherwise the field can spontaneously reverse and the data is lost.
So one day the atoms might just realize that they've been tricked and you'll end up with your computer on fire because your benzene chains have all broken and you end up with
2-methyl-1,3,5-trinitrobenzene
Yeah, and not to mention that cobalt atoms can be very nasty [wikipedia.org] if they decide to isotope themselves. Will the EPA to allow them in PCs?
Early beta tst (Score:5, Funny)
I trhed an e"rlx be|a tast.( Uhe res7ltw w%ren/t so pretpyn
Useless if the speed is the same (Score:2)
Well this is all fine and dandy for storage space, but what about performance? Drives are getting bigger all the time, but you're still stuck with spindles that rotate at the same speeds as the ones from last year, or the year before. I can't see anyone wanting to replace their speedy many-spindle database disk farm with a single 320TB disk that still spins at 10Krpm and only delivers ~125 IOPS. Performance is going to suck big time. All the top TPS benchmark results for example are achieved using 100
Re: (Score:2, Interesting)
As data density increases, so does the rate at which it can be read. Assuming two orders of magnitude increase (100x) and individual bits staying roughly the same shape, the linear density increases by a single order of magnitude. (10x bits per track, 10x tracks). The drive will be able to read at 10x the speed.
At 3 orders of magnitude, you can expect a read speed improvement of roughly 3000%. (sqrt(1000) ~31.6)
Re: (Score:3, Interesting)
Ok sequential IO is going to improve some as more data will pass under the head compared to todays disks. But Random IO isn't going to feel the same benefit because that's influenced more by Rotational Delay (fixed by spin speed) and the time it takes for the head to shift between tracks: Disk Seek. So your figures are going to be wildly off in real life.
Re: (Score:2)
First, random IO is mainly dependent on seek time, not rotation speed.
But we try to not do random IO at disks anyway, currently people are quite good at avoiding it.
Re: (Score:2)
Re:Useless if the speed is the same (Score:5, Interesting)
Back when I was in college one of the 'cool' old Comp Sci professors had a tale he liked to share with his classes on the first day. I had him in a couple of classes, so I heard it over and over again. His presentation made it an amusing story if you could get over the fact that he smelt as if he lived in an ashtray.
It seems that back in the mainframe days, the standard way of increasing storage size on your hard drives was to make a bigger platter. Seems rather simple, right? The storage size grows exponentially with its radius. So adding an inch each time can lead to some fairly nice results, and with some platters topping out at 24 inches, that's some space.
Except....
One day, the university ordered the 'latest' hard drive for one of their mainframes. I'm sure it was a behemoth, it probably held around 50 meg. The vendor came by and installed it, and everything seemed fine till a few months later when the drive seemed to start failing, at about 30% capacity, writes stopped working and anything written to seemed to have been corrupted. They were puzzled, but this is why such things service contracts. The vendor came out, replaced the drive, and everyone went on with life.
Till it happened again, at about the same capacity. Another replacement was made and vendor was quite red-faced and explained that they seemed to have run into a batch of dud drives. All was forgiven and life went on.
Till, it happened the third time. At this point, it was starting to embarrass everyone: The vendor, the people who ordered the hard drive in the first place, etc. So this time, instead of just allowing the vendor to take the drive back, the dean of the department demanded they diagnose the issue there on the spot.
Now, this wasn't the age of the sealed drive cases, certainly drives were still kept 'clean' but we weren't to the point yet where a single grain of dust could wipe out megabytes of info (heck, even the 24 inch platters needed to be in arrays of 50+ just dream of hitting 100 meg) so cracking open the drive wasn't that big of a deal.
So the vendor's tech, hoping to appease a clearly angry customer in the day and age when parts cost tens of thousands of dollars, popped open the drive.
Want to guess what they found?
Larger disks do indeed result in more surface area, but they also result in a higher centrifugal force on the edges. An increased force which the vendor apparently hadn't accounted for. Once the disks began to spin up, the glue holding the magnetic dust to the platter gave way, resulting in the platters being stripped clean after a certain radial length from the center. The disks themselves were fine up to that point, the dust was plastered to the case itself and when the platters came up to speed any dust that had fallen back onto them was once again flung up against the case.
The reason why the disks didn't seem to fail till they reached a certain capacity was simply because they weren't being used in a RAM fashion but were being written to in a sequential manner. The outer portions of the platters were only being hit once the inner portions were written to.
Perhaps the reason spindle speeds haven't gone up lately could be part of the same issue. Or perhaps I'm simply indulging in a bit of pointless nostalgia as I wait for this report I'm running to finish. Who knows...?
Parent
Re: (Score:2, Informative)
The storage size grows exponentially with its radius.
At a fixed data density (and a fixed number of platters), storage is proportional to the area of the platter, which is proportional to the square of the radius.
Storage size grows quadratically with its radius, not exponentially.
Math? (Score:2)
OK, what am I missing here? 0.5nm is 16 times smaller than 8nm. On a 2D platter, that's 256 times more dense, not 1000 times more dense.
-S
Re: (Score:2)
Not that it's necessarily the case, but I'd imagine there is some minimal spacing dictated by the strength of the magnetic fields in use. The smaller structures could allow for smaller spacing than would be allowable for their contemporaries.
Experiment? (Score:3, Insightful)
sounds dangerous (Score:2)
Benzene [wikipedia.org]
Re: (Score:2)
Sure, but if it were to somehow leave the device, that would mean there are cells that can no longer function. I can't imagine this could make it in the market if any appreciable amount benzene could be "lost" after manufacturing.
Though, if my math is correct, you only need about 1 picogram of Benzene for a Gigabyte worth of cells. Not worth considering even if it wasn't locked away inside a silicon and ceramic package...
Re: (Score:2)
isnt benzene a carcinogen?
Not to worry -- we will put a big red sticker on the side of the drive that says "DO NOT EAT".
call me stupid (Score:2)
1000x Denser ???
... 50,000 atoms ... to 15,000 atoms... 8nm across ... only 0.5nm across... could allow an increase in memory density of three orders of magnitude
three orders of magnitude? what kind of math is this???
The only question now is whether this team's calculations hold true in the real world.
I would like to see that calculation!
AI? (Score:2)
This may be the breakthrough, though, that allows for the type of density that would be required for a human-analog type AI to be a reality.(currently it would take a small building to approximate a typical human brain)
Re: (Score:2)
OK, just stop right there. Porn and beehives do NOT mix.
Re:1000x denser (Score:5, Funny)
OK, just stop right there. Porn and beehives do NOT mix.
Rule 34.
Parent
Re: (Score:2, Funny)
Bee Porn [youtube.com]
Re: (Score:2)