New implementation of graphene based technology could multiply the storage density tenfold on HDD's

[Fnige]

15 hours ago, LAwLz said:

There are so many "revolutionary" discoveries that promise major breakthroughs that just end up being impractical for various reasons. 

Or always 10 years away from being actually used instead of tested

[Lightwreather]

15 hours ago, SeriousDad69 said:

A 200TB HDD would be worthless unless they also found a way to make them run like 8X faster lol. Picture trying to fill/drain a swimming pool with a straw. Even 20TB HDDs are too big for their current speed imo

I mean, they are soon getting support for the Nvme standard, then there's a quite a bit of other stuff going on to make the faster

[leclod]

Edit : don't bother

[Roswell]

Oh look... Another graphene breakthrough that's apparently going to be awesome but never materializes like the other bajillion graphene things.

 

Beyond that, a 200-300TB mechanical drive would take days to fill in a server environment. I don't think 5 days of continuous pedal to the metal mechanical disk writing would be super great for reliability.

 

 

[leadeater]

6 hours ago, J-from-Nucleon said:

I mean, they are soon getting support for the Nvme standard, then there's a quite a bit of other stuff going on to make the faster

Interface speed and protocol isn't a speed issue when it comes to HDDs. SAS-3 (12Gb, 1.5GB/s) and SAS-4 (24Gb, 3 GB/s) already exist and SAS support better command queuing and simultaneous read and write operations and putting either interface on a HDD that can only sustain 150-200 MB/s won't make it any faster.

 

NVMe making it's way to HDDs has little to do with speed and more to do with common standardization and protocol features like NVMeoF.

 

21 hours ago, SeriousDad69 said:

A 200TB HDD would be worthless unless they also found a way to make them run like 8X faster lol. Picture trying to fill/drain a swimming pool with a straw. Even 20TB HDDs are too big for their current speed imo

When density increases so does throughput, how much isn't really known as it depending on recording technology. You can see this in action with LTO tape standards though, with up to one to one doubling of density to doubling of speed though not always, the it depends factor.

[Moonzy]

On 6/14/2021 at 6:41 PM, Lurick said:

It's so revolutionary bro!!

because hard drives

[Lurick]

1 minute ago, Moonzy said:

because hard drives

I honestly didn't even think about that, lol

[lostcattears]

So how expensive it will be to make one of these? 

[CarlBar]

20 hours ago, Roswell said:

Oh look... Another graphene breakthrough that's apparently going to be awesome but never materializes like the other bajillion graphene things.

 

 

As soon as the figure out a way to do graphene in a cost efficient manner expect a lot of them to suddenly pop up in actual use. In many ways stuff like this creates more push to find cheaper ways to work with graphene or make graphene.

[RejZoR]

Do this. Then do the dual actuator design used by Seagate Exos drives (MACH.2) on each actuator. It should have roughly 1GB/s sequential throughput per drive then. I'm also fairly certain they can make it more compact than this Conner's design from the 90's. Ironically, Seagate bought Conner so the MACH.2 drives are sort of homage to the old times.

[GhostRoadieBL]

So say you increase the storage density 10 fold, seek time increases by the same factor or you need to increase platter speed to compensate, more cooling, more air resistance etc. The volume of data to be read from a full drive would outway the gains from the density per drive.

 

Even if that advancement happened tomorrow, and then 200TB drives are available, the space vs storage ratio still favors nvme drives. How many 8TB nvme sticks can you fit in a 3.5in drive? Let alone just chip stacks on a single controller.

[leadeater]

10 hours ago, GhostRoadieBL said:

seek time increases by the same factor or you need to increase platter speed to compensate

Seek time would not increase by the same factor. Seek time is related to the rotational speed and the time it takes to lock on to a track and then start reading the required sector. If you double the density and nothing else changes then depending on how the density was doubled then throughput is potentially doubled.

 

If we keep it to more simplistic/traditional recording format technology, CMR/PMR, then increasing the density general worst case slightly increases the track lock time as the actuator has to be more precise which could take longer.

 

This is how HDDs have actually gotten faster over the years, platter density increases which increases throughput.

 

Seagate ES (2013 era technology)

 

 

Sustained transfer rate 62 to 152 MB/s

 

Seagate EXOS X16 (2019 era technology)

 

Sustained transfer rate 103 to 249 MiB/s

 

More than 3 times the density, quite a lot better throughput. Couldn't find access times for the EXOS disks but I know it's not applicably different or potentially better, maybe. These newer disks have much better sustained IOPs but that may largely come just from the density increase and not access time improvement. Access times for CMR/PMR hasn't changed much in a very long time.

 

Where access times might see some appreciable difference is in HAMR recording technology has you have to heat the platter/sector you are going to write to, read times far as I know should be the same.

[Jet_ski]

Did they learn nothing from this? Graphite+heat source = water cooled hard drives => catastrophic disaster

[TOMPPIX]

Could a hard drive have 3 or more actuators?

[cj09beira]

On 6/16/2021 at 9:13 AM, RejZoR said:

Do this. Then do the dual actuator design used by Seagate Exos drives (MACH.2) on each actuator. It should have roughly 1GB/s sequential throughput per drive then. I'm also fairly certain they can make it more compact than this Conner's design from the 90's. Ironically, Seagate bought Conner so the MACH.2 drives are sort of homage to the old times.

the current 3.5 inchx5.78inch is too small to be able to have 2 actuators but i think they could do it by simply increasing the length a bit to around 6.5 inches.

if i was running a hdd company i would have put lots of pressure for the industry to go in that direction, increase the size of the disk a bit and double performance doesn't seem that hard of a thing to pull off, the stacking of actuators that seagate is doing seems much harder

1 hour ago, TOMPPIX said:

Could a hard drive have 3 or more actuators?

over 2 is really not pratical, unless they are stacking the actuators then sure the problem is just how to have more magnets between the actuators that are strong enough for the job

[Loote]

They do have multi-actuator in Mach 2 tech, but what you've shown in a pic would be even better.

So, the total time it takes to read a HDD of whatever size is just how many circles of data there are from the middle to the edge of the platter, divided by rotations per time unit and adjusted for the aforementioned time to lock on a track, a drive with 7200 rpm, 6000 'circles' and 1.(6)ms to lock would take 8.(3) ms to complete a circle, ~10ms to complete a circle and switch to another one, which rounds up to a nice 100 circles/second and a minute to read the whole drive(I pulled numbers from my butt to get this result).

Now, a circle will have different circumference depending on how far it is from the center, but let's ignore it and go with an average. Let's even call it x cm because I don't care to check how big the platters are. Now they say they can fit y bytes per unit of length or surface, that influences how much data fits in that average circle, but also how many circles there are, 10% denser is 10% more bytes/circle, but also 10% more circles/drive, which is why HDDs are faster at growing in size than in speed. If life was so easy getting 9x the size of a drive means 3x the circles and 3x the amount of data per rotation, so 3x as fast sequential speed, equal seek time and 3x longer to write the entire drive.

Now we know Mach 2 nearly doubles the speed, the tech from the above post could do that once more, meaning 200TB drive with 4 actuators could take the same time as single actuator 20TB drive today. There is of course question of what clients need, not every use case cares greatly about taking 4 days to write an entire drive and would gladly take 4 days and have cheaper storage than pay more for the speed they don't need. For example I have collected 50TB of data over the years, at once I move maybe 200GB of it, Taking 5 days to write everything I have onto a single drive that I could keep in a different physical location is not an issue at all, many archives are just like that, 1TB copy but you could fit 200 of those on 1 drive, getting that 1TB copy would take a bit over an hour in case of a drive that is 200TB and takes 5 days to write in entirety(depending on position on the platter probably more like from an hour to two and a half), for many that is acceptable.

 

Spoiler

This is a really slow external 2.5' 5TB SMR drive:

The q1 write speeds are especially painful
I feel like this is what people base their opinions of HDDs on, meanwhile here's the cheapest external 10TB drive I found year ago:

it can read 100GB in 10 minutes, not great, not terrible.  It is CMR and 3.5' though.
Should be quite clear why drives with dual actuators and higher densities need to go beyond SATA 6gbit.

Would be pretty funny if SATA Express made a comeback now, it had 2 PCI-e lanes in 2.0 times, though that's never coming back I think.