Seagate still HAMRing away at the 100 TB disk drive decades later

Seagate says it has a clear way forward to 100 TB disk drives using 10 TB per platter technology, but HAMR tech is nearly 25 years old and full mass production is still not underway. What has been taking so long? HAMR (heat-assisted magnetic recording) technology is used to write data to a granular iron platinum medium that …

[Anonymous Coward]

Thanks for the MAMRies

I used to work at HGST before WD fully Borg'd them.

My main role was in Enterprise class flash storage and Petabyte-size object stores and for the latter I was on the internal R&D mailing list.

I remember the day the MAMR folks sent out an exciting bulletin about the drive tech. Words to the effect...

Good news, we've stopped the heads from exploding.

Which was nice.

[Lee D]

Re: Thanks for the MAMRies

In my experience of Seagate drives, that would be an improvement because they actually STOPPED the heads exploding.

I'm not sure I've ever managed a Seagate drive that DIDN'T fail. And I can't say that about any other manufacturer (e.g. HGST, WD, even Maxstor and IBM back in the day).

[Blank Reg]

Re: Thanks for the MAMRies

Same here, I haven't had a huge number of dead drives, but every dead drive was a Seagate. And it's all the worse because I primarily buy anything other than Seagate

[Hubert Cumberdale]

Re: Thanks for the MAMRies

Not my experience. Seagate has always been fine for me. I have a theory that everyone sticks with a drive manufacturer until they get a failure (or a cluster of failures; see Poisson distribution and bathtub curve), at which point they decry <manufacturer> as being useless, never to buy from them again. I've heard the same "I've never had a <manufacturer> drive that didn't fail", or "never buy a <manufacturer> drive" about pretty much all manufacturers.

Having said that, looking at Backblaze's recent statistics of annualized failure rates vs. manufacturer (graph about 2/3rds of the way down the page), Seagate does seem to have been above the average for a while. But it's most recently almost at the average and far exceeded by HGST. Western Digital seems to be the current best bet. But I think that the huge variability over time per manufacturer still gives my point validity: they all go through better and worse patches.

[Alan Brown]

Re: Thanks for the MAMRies

The bigger problem with failed drives is how the warranty claims go (badly in my experience)

[localzuk]

HAMR

The Duke Nukem Forever of hard drive tech.

[Anonymous Coward]

Re: HAMR

Yeah, and, to be honest, "Cook your hard disk with a fricking laser beam" definitely sounds like "Oh, what could possibly go wrong" technology…

"Sorry, we can't read your file, your data sort of melted and smeared itself around a bit…"

[NoneSuch]

Re: HAMR

– but only when the bit areas are heated using a laser.

Should be:

– but only when the bit areas are heated using a "laser."

Additional 'finger quotes' are just not optional with that word. Now, if you'll forgive me, I need to find sharks with frikkin laser beams strapped to their heads.

[ecofeco]

25 years?

That ship has sailed. No doubt one day the bottleneck will be solved, but by then, the world will have moved on.

[Annihilator]

Re: 25 years?

I always felt like it was going to be a race between HAMR and holographic storage. But then solid state arrived and became cheap enough...

[Eecahmap]

Re: 25 years?

I still want Harry Harrison's molecular memory.

[Stevie]

Bah and Double Bah (BDB)!

I tried reading the article (RTA) from top to bottom (TTB) but found the confusing multitudinous proliferation (CMP) of three-letter acronyms (TLA) - and yes I know that they aren't really acronyms (ARA) if they don't spell out words (SOW) - to be detrimental (TBD) to reading the article (RTA) for comprehension and retention (CAR) of information.

No doubt the technology is awesome (TIA) but I CBA to read the article.

[DJO]

Re: Bah and Double Bah (BDB)!

I understand where you are coming from but every abbreviation was explained in the text:

HAMR (heat-assisted magnetic recording)

perpendicular magnetic recording (PMR)

shingled magnetic recording (SMR)

cloud service providers (CSPs)

energy-assisted PMR (ePMR)

By all means complain with justification where they are not explained but this article was not guilty in that respect.

[Stevie]

Re: Bah and Double Bah (BDB)!

In these days of painless copy and paste one wonders why we need to resort to such lengths in the first place.

Especially if one is confronted by so many of the things one forgets what one means and has to wade through the whole article to rediscover the expansion.

[Roland6]

Re: Bah and Double Bah (BDB)!

The use of abbreviations, probably helps to make the article less useful as AI training data. Plus for the human reader reduces the verbiage and length.

[Eecahmap]

Re: Bah and Double Bah (BDB)!

A CDR for every CAR. . . .

[ExampleOne]

60TB flash SSDs are already available at not insane (for commercial use) costs, 120TB is announced (though I haven't found a supplier actually offering them).

I'm not sure how interesting 100TB HDDs are really going to be in 5 years time.

[Kevin McMurtrie]

The problem with high density flash storage is that it gradually fades away. It's not good for long-term storage without power. Hard drives can fade too but it's slower.

[DS999]

I don't think anyone is using hard drives to store data long term without power - that's where tape shines (and is another 10x cheaper than hard drives) and there's no worry about your hard drive deciding not to spin up if its been sitting on a shelf for a few years.

What hard drives do allow is to conserve power by spinning them down when the data they contain is not immediately needed - nearline availability. Tape can do that too but the latency to the data from a tape library is another order of magnitude or two longer than the time it takes a hard drive to spin up. Plus it requires all the data you want is essentially sequential, where a hard drive allows it to be accessed randomly across one or multiple hard drives.

SSDs could be operated the same for long term low power storage - they'd just need to be powered up occasionally, sort of like DRAM refresh cycles. I don't know if simply powering them up for a few seconds once a day would be sufficient, or if they would eventually need to have data rewritten, but you wouldn't really have to worry about write wearout if you were only rewriting say once every few months. Anecdotally I've had thumb drives sit in a drawer for several years and still have the data I wrote to them. They are too old to be QLC and maybe not even TLC but they sure weren't using the cream of the crop NAND technology either.

No, the reason people aren't using SSDs for that nearline availability that hard drives are now fulfilling is down to one thing. Price.

[Alan Brown]

For all intents and purposes, flash outlasts HDDs by quite a bit

Neither HDD or flash are qualified or recommended for long term storage. You use them for this purpose at your own risk

[Androgynous Cow Herd]

depends on the use case

Sure, yes, 60TB SSDs are available now.

They also are over 10x the cost of spinners on a per TB basis.

The need for large geometry spinners is that a lot of large data sets don't need SSD level latency, and most enterprises have these pesky things called "Budgets"

For Every dollar you spend on SSD capacity, you can spend a dime for same capacity on spinners.

Or, you can spend a penny on tape media.

Bandwidth is cheap. Latency is expensive.

(All that said, I wouldn't touch a Seagate shingled drive with a 10 foot stylus, and HAMR seems like a really bad idea)

[ExampleOne]

Re: depends on the use case

The current price for spinners is around £20/TB for the largest ones. The price for a 60TB SSD is around £6000 or £100/TB. That is current prices for those SSDs.

Give it another 5 years and presumably they will drop in price. I am not convinced HDD prices will drop as fast, narrowing the price gap. At some point SSDs will overtake spinners on the price metrics.

And yes, I agree tape will continue. The problem for spinners is their niche is slowly shrinking as cheaper SSDs narrow the gap to tape.

[DS999]

Re: depends on the use case

I am not convinced HDD prices will drop as fast

They won't, simply because the niche for hard drives is growing smaller by the year. A lot of R&D was being invested in hard drives when that was the one technology, and even when SSDs appeared there was plenty of money pouring into hard drive development when they were still a cheaper alternative for low end PCs.

We're now at the point where it really doesn't make financial sense to use hard drives even in the lowest end PCs. Some may still do but that would likely be the OEMs having old stock of hard drives or opportunistically buying someone else's old stock for pennies on the dollar, so none of that would be new money flowing into the coffers of Seagate, etc.

So now their ONLY market is bulk storage of data at lower prices per bit than SSDs. How much longer will Seagate deem it worthy to keep developing HAMR technology to allow these future 100 TB drives? At some point they will probably say "this future R&D isn't going to pay off, let's just optimize the tech we have now and try to drive down the manufacturing cost" and future capacity increases will slow a crawl or possibly even revert if that's the tradeoff they get for significantly lowering manufacturing cost. I would bet heavily against ever seeing 100 TB or even 60 TB hard drives. I think the beancounters stop the R&D money long before they get there.

Once they've squeezed everything they can out of their manufacturing cost it'll no longer be a moving target for SSDs, and once SSDs get to say 2x the hard drive per bit cost then they probably announce a last order date a year out and that's the end of the hard drive.

[Alan Brown]

Re: depends on the use case

"They also are over 10x the cost of spinners on a per TB basis."

Comparing SSD and HDD prices is an apples to oranges thing

Factoring in longevity and power, the jumping off point for SSD vs HDD is 4-5x

[DS999]

Re: depends on the use case

Not for bulk storage. Power isn't a factor because you only spin up drives when they're needed, and longevity is mainly governed by depreciation cycles for the enterprise market (the only remaining hard drive customer that matters) so longevity beyond that is irrelevant.

I think 2x is the mark to hit to kill hard drives. That's when you see whoever is the last one standing in hard drives (presumably Seagate) announce a last order date.

[Tubz]

Quantum Bigfoot 2.1GB most reliable drive I have and still going in my retro 486 machine, when Seagate can provide that type of service, I may buy another one of their drives or hell freezing over.

[Alan Brown]

You're extremely lucky. Very few Bigfoots lasted more than 5 years. They were notoriously fragile