Next-Gen HAMR Platters Promise 80TB Hard Drives

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,785
"Conventional magnetic recording methods today achieve 1.14Tb per square inch, but SDK believes its HAMR technology will increase that to up to 6Tb per square inch. Based on today's nine platter hard drives, that means an 80TB hard drive is theoretically possible. However, it's not clear what storage density SDK has managed to achieve so far. So while we will get to 80TB drives, we don't know when yet."

https://www.pcmag.com/news/next-gen-hamr-platters-promise-80tb-hard-drives
 
(also from Feb 7th funnily enough) https://arstechnica.com/gadgets/202...aser-assisted-hard-drives-are-coming-in-2020/

Freakin laser-beams in your hard drives! LASERS!
Hmm

1601398720977.png
 
Well with storage requirements what they are for big data centres I’m not surprised we’re getting to this point but sweet Jebus.... the backups.
 
Yes that's becoming a real problem of actually getting data on and off of these drives. There is always some gain in sequential read / write speeds because of density. More bits will pass under the head in one rotation of the drive. The flip side of that coin is that there are more tracks every time you can shrink the area, so it requires an additional rotation for each new track to be read. At a certain point it will simply become infeasible to continue to increase storage density unless you can pick up multiple tracks at one time. These drives will be fine for archiving where you are only ever going to write to them once and maybe never even read them again, but they simply can't keep up with the increasing I/O demand. (Read, good for writing video footage to, bad for playing back multiple video streams at once)
 
cool but everything getting smaller, so still waiting for nvme drives to become that size.
Prepare to wait 5-10 years then, because the historical data and trendlines tell a different tale, as NAND doesn't just linearly or infinitely scale. If anything, spinning disks have been making a resurgence, as SSD sizes have hit a bit of a wall.

People have been blindly assuming the demise of spinners for the past 10-15 years, and they're only getting bigger and more expensive/TB.
 
Seagate has been working on HDDs with the set of read/write heads split in half so that it can access 2 platters at once for a few years (Google showed results back to '17 in my search). Theoretically they could go as far an independent head for each platter.

https://blog.seagate.com/craftsman-ship/multi-actuator-technology-a-new-performance-breakthrough/

Yea it's still a bandaid fix though. I think there were some test drives out there that had dual heads that covered the same platters, so you could double up speeds on one platter. That would at least someone address the I/O problem, but it just resets the issue by half. Double the size of the density and you're back to where you started. The Seagate one isn't anything special IMO, because you can achieve the same thing by just buying 2 8TB drives instead of 1 16TB one. Either one of those drives still have the issue with increasing number of tracks, and if the data you're looking for is random and was written onto the top half of the drive in different tracks, having a separate head for the bottom half didn't gain you much. Basically it's useless at low queue depths, because it's unlikely that you'd have enough randomness to the access to make a difference. I can see at high queue depths were it could end up being more than a 2x benefit, but the cost of actually implementing it you can probably just pick up and SSD and destroy one of these in random reads.
 
The tech is interesting for sure. If a large spinner dropped on the market at a good price id consider purchasing it just for bulk storage. Less wear on my nvme storage.

Right now i have 3 4tb spinners, be nice to get rid of two drives worth of sata data/power cables.

For my collection of video clips. :) Dungeon and Dragons characters and art, and all data where slow is not an issue.
 
Yea it's still a bandaid fix though. I think there were some test drives out there that had dual heads that covered the same platters, so you could double up speeds on one platter. That would at least someone address the I/O problem, but it just resets the issue by half. Double the size of the density and you're back to where you started. The Seagate one isn't anything special IMO, because you can achieve the same thing by just buying 2 8TB drives instead of 1 16TB one. Either one of those drives still have the issue with increasing number of tracks, and if the data you're looking for is random and was written onto the top half of the drive in different tracks, having a separate head for the bottom half didn't gain you much. Basically it's useless at low queue depths, because it's unlikely that you'd have enough randomness to the access to make a difference. I can see at high queue depths were it could end up being more than a 2x benefit, but the cost of actually implementing it you can probably just pick up and SSD and destroy one of these in random reads.

For data center purposes the limitation shouldn't be serious in practice. You'd mostly be barfing near-read only archival data into them; and like with SMR with storage appliances that are aware of how it works and its limitations feeding data to both sets of heads at once shouldn't be an issue. SMR is far more obnoxious in this regard, and unlike on consumer devices where it's a WTF drenched in failsauce, high end enterprise SANs and their cloudy equivalents can made good use of SMR drives.

Making it work for consumer drives would be trickier; but unlike SMR might be doable by moving the magic to the drive firmware. eg an even number of platters and splitting each sector in half so the first 2048 bytes are in a top platter, ad the second 2048 a bottom one.
 
Last edited:
Prepare to wait 5-10 years then, because the historical data and trendlines tell a different tale, as NAND doesn't just linearly or infinitely scale. If anything, spinning disks have been making a resurgence, as SSD sizes have hit a bit of a wall.

People have been blindly assuming the demise of spinners for the past 10-15 years, and they're only getting bigger and more expensive/TB.
Which wall is that? The 100TB wall? You do realize you can buy larger SSD's than HDD's right? https://timesofindia.indiatimes.com...gest-ssd-costs-40000/articleshow/76878101.cms
Ok, it's cheating it's a 3.5" drive! We are talking about a 20TB HDD coming out when there are already 100TB SSD's talking about how SSD's are hitting their limits. Seems a bit backasswards. Sure, the price isn't something most would pay, but just shows we're not anywhere close to the "limit/wall" for SSDs.

ps. It comes out to about $400 per TB, which really isn't horrible for a server drive. Most high quality 1TB SSD's are going for 200+ per TB. Server quality are much more, an example is a dell 7.68TB drive, which sells for $25,000 which comes out to $3,255 per TB... so $400/TB is much closer to consumer prices than server prices.

https://www.dell.com/en-us/shop/del...-pwjP5L_d8xeREZOmrAaAmp5EALw_wcB&gclsrc=aw.ds
 
Which wall is that? The 100TB wall? You do realize you can buy larger SSD's than HDD's right? https://timesofindia.indiatimes.com...gest-ssd-costs-40000/articleshow/76878101.cms
Ok, it's cheating it's a 3.5" drive! We are talking about a 20TB HDD coming out when there are already 100TB SSD's talking about how SSD's are hitting their limits. Seems a bit backasswards. Sure, the price isn't something most would pay, but just shows we're not anywhere close to the "limit/wall" for SSDs.

ps. It comes out to about $400 per TB, which really isn't horrible for a server drive. Most high quality 1TB SSD's are going for 200+ per TB. Server quality are much more, an example is a dell 7.68TB drive, which sells for $25,000 which comes out to $3,255 per TB... so $400/TB is much closer to consumer prices than server prices.

https://www.dell.com/en-us/shop/del...-pwjP5L_d8xeREZOmrAaAmp5EALw_wcB&gclsrc=aw.ds
A bunch of drives glued together into a single "drive" that costs 5 or 6 digits for the industrial/enterprise space isn't what I was talking about.

Capacity growth in terms of form factor and price/TB trendlines for the consumer space - meaning a single 2.5" or M.2 stick - we've hit a wall at 4TB where it's not really getting cheaper nor are ever-larger form factors coming out as frequently as they were before. And certainly not to the extent I've been hearing people assuming for years - that "no point in buying spinning disks cuz SSD will just take over". It was naive 5 and 10 years ago, it's naive now.
 
Last edited:
A bunch of drives glued together into a single "drive" that costs 5 or 6 digits for the industrial/enterprise space isn't what I was talking about.

Capacity growth in terms of form factor and price/TB trendlines for the consumer space - meaning a single 2.5" or M.2 stick - we've hit a wall at 4TB where it's not really getting cheaper nor are ever-larger form factors coming out as frequently as they were before. And certainly not to the extent I've been hearing people assuming for years - that "no point in buying spinning disks cuz SSD will just take over". It was naive 5 and 10 years ago, it's naive now.
I'm not sure that I would consider a 3.5" drive that crazy of a design... I mean, it's the size of a standard HDD that we're comparing it to. Also, as far as SSDs are concerned, the pricing isn't even that horrible. $400/TB is a bit high, but you are going to pay for a large capacity drive. My point was SSD's haven't hit a wall, so much they are still more expensive than HDDs per TB and people are only willing to spend so much.
Heck, you can pick up a 2.5" samsumg 30TB SSD for a cool $8k, which comes out to $266 per TB, which is really getting close to reasonable (per TB reasonable), even if the total cost seems unreasonable. This is a server rated 12Gbs SAS drive... so again, SSD's can be bought in much larger sizes than the "wall" you're talking about. Sure people will spend $400 on a 2TB drive... but they won't spend 8k on a 30TB drive. The cost per TB isn't that far off, it's just people aren't spending that much on a drive. The only wall they ran into is a COST wall, not a capacity wall.


Edit: Forgot to link the samsung, just incase someone is looking for a few for their RAID ;).
https://www.serversupply.com/SSD/SA...REH2gu-Vg4b7sErFQJepTDIQlVt6BFxkaAoH8EALw_wcB
 
"Conventional magnetic recording methods today achieve 1.14Tb per square inch, but SDK believes its HAMR technology will increase that to up to 6Tb per square inch. Based on today's nine platter hard drives, that means an 80TB hard drive is theoretically possible. However, it's not clear what storage density SDK has managed to achieve so far. So while we will get to 80TB drives, we don't know when yet."

https://www.pcmag.com/news/next-gen-hamr-platters-promise-80tb-hard-drives
Don’t forget power draw on SSD’s becomes an issue you could cram as many as you can fit into a form factor but you have to power it with something. Platters you just build a motor capable of the desired speeds for the power allotment. 80TB of HDD pulls a lot less juice than 80TB of Solid state does.
 
We need to go back to the 5.25" form factor, the bigfoot needs to return.

View attachment 284166

We have all of these 5.25" bays and nothing to put in them. Let's marry old ideas with new tech and have even larger storage drives.

LOL -- I just rebuilt my system -need to update my sig - and was looking at those very 5 1/4 bays of which you speak and thinking what to do with them.
 
The 20TB does not overly impress me given we already have 18TB drives. I expect these huge ones are still years away. Also are these going to be enterprise only drives with enterprise pricing?
 
We need to go back to the 5.25" form factor, the bigfoot needs to return.

View attachment 284166

We have all of these 5.25" bays and nothing to put in them. Let's marry old ideas with new tech and have even larger storage drives.
Big foot is nothing. Seagate Elite 9 was 5.25" Full Height.
1601526834279.png
Image pulled from EBay seller. 14 platters, 28 heads (27 for data, one for servo control info). I got four of these in ~ 2003, they were supposed to be old new stock, but they were pulls with tons of hours (and one wouldn't spin up without help) and I got my money back. Loud as hell, took about two minutes to spin up. I bet you could fit a lot more than 14 platters in that space now, the article says they're putting 9 platters in current drives (and I see that in drive specs for top models)
 
Last edited:
Big foot is nothing. Seagate Elite 9 was 5.25" Full Height.
View attachment 284481
Image pulled from EBay seller. 14 platters, 28 heads (27 for data, one for servo control info). I got four of these in ~ 2003, they were supposed to be old new stock, but they were pulls with tons of hours (and one wouldn't spin up without help) and I got my money back. Loud as hell, took about two minutes to spin up. I bet you could fit a lot more than 14 platters in that space now, the article says they're putting 9 platters in current drives (and I see that in drive specs for top models)
I can only imagine the power draw on that thing.
 
I bet you could fit a lot more than 14 platters in that space now

The platters were pretty thick on these old drives. Also they did not spin so fast.
 
Big foot is nothing. Seagate Elite 9 was 5.25" Full Height.
View attachment 284481
Image pulled from EBay seller. 14 platters, 28 heads (27 for data, one for servo control info). I got four of these in ~ 2003, they were supposed to be old new stock, but they were pulls with tons of hours (and one wouldn't spin up without help) and I got my money back. Loud as hell, took about two minutes to spin up. I bet you could fit a lot more than 14 platters in that space now, the article says they're putting 9 platters in current drives (and I see that in drive specs for top models)

I have a 1988 760 MB Micropolis SCSI drive which is that big. Still works and store some data on it.
 
Was it actually usable/fast in practice?
For the most part it was noticeably faster, but definitely not the advertised speed they where saying. I believe It only lasted me a few years before it started to not work.
 
Can you fit more then one word/excel doc on it? :)

I've got a spreadsheet on my work computer that is too big to fit on it. An abomination that weights in at about 1GB on disk and takes a half dozen times that much memory to open.

I had a problem, so I decided to use regular expressions Excel. Now I have two problems.
 
Back
Top