HAMR Hard Disk Drives Set To Arrive In 2018

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
Here comes yet another hopeful estimate as to when we may finally get HAMR in our hands.

Numerous demonstrations of HAMR-based HDDs in action prove that the technology actually works. Over the years, producers of hard drives, platters and recording heads have revealed various possible timeframes for commercial availability of drives with HAMR technology. Their predictions were not accurate. At present, there are still reliability issues with the technology, according to Seagate. In the recent months both Seagate and Showa Denko indicated that HAMR drives would be delayed again.
 
Watching hard drives grow over the years has been amazing. I remember my first 100MB drive, I thought I would never fill it full. Then, I remember my first 1GB drive, amazed at what seemed like an bottomless pit of storage. Now, I'm running 2TB and need more than ever.

This tech seems neat and all, but, I wonder what can be done if they move away from the standard, and re-invent the hard drive. There has to be something better out there than spinning metal disks!
 
Is this actually going to be a viable pursuit considering Xpoint and Memristor SSD tech on the horizon?
 
Hard drives are obsolete dinosaur technology set to be replaced even in mass storage by gigantic SSDs in the next few years. At this point, who cares?
 
Hard drives are obsolete dinosaur technology set to be replaced even in mass storage by gigantic SSDs in the next few years. At this point, who cares?

In a word, cost. Solid state drives still have a ways to go in cost to truly obsolete hard drives.
 
Yup, think outside the desktop market when your need hundreds of terrabytes of storage...SSD now are a huge cost vs spinning rust.
 
Hard drives are obsolete dinosaur technology set to be replaced even in mass storage by gigantic SSDs in the next few years. At this point, who cares?

Not with NAND Flash they aren't. NAND Flash isn't even long-term storage, with data degradation with TLC Flash starting after just a few months. There's also no way at this point to make NAND Flash cheap enough to directly compete with HDDs. The latter will remain 5-10x cheaper per GB for the foreseeable future.

XPoint storage/memory is the only real competition on the horizon for HDDs, and I don't see that one becoming a major threat for the first couple of years.
 
Hard drives are obsolete dinosaur technology set to be replaced even in mass storage by gigantic SSDs in the next few years. At this point, who cares?

Spinning disk will for the foreseeable future remain less expensive per TB than SSDs. In our environment we have roughly 8,000 spinning disks in production. Some of that will be replaced by flash storage which comes with compression and deduplication.

A 20 TB flash storage unit runs about $10,000/TB (EMC XtremIO) so even with compression and dedup makes that 20 TB into say 60 TB effective (that's optimistic), it comes out to ~$3,300/TB and that's a far cry from the less than $500/TB for spinning SAS disk (not just counting the disk, but the cost of the controller etc. as well).
 
In a word, cost. Solid state drives still have a ways to go in cost to truly obsolete hard drives.

Certainly, but next generation SSDs are getting cheaper while next gen HDDs are getting more expensive, so the gap is closing faster than before.

Plus it seems SSDs will outclass HDDs in storage capacity too. 16TB SSDs could come as early as next year while hdds will top at 10TB. HAMR should boost to 100TB but by then SSDs should be past 500TB
 
I want to know what the performance will be like. I have to imagine that the heating trickery used to squeeze in more data is going to increase read and write times. SMR drives are already a big step backwards.
 
^ Sure, you could buy one of those large new-fangled SSDs if you won the lottery ;)
I want to know what the performance will be like. I have to imagine that the heating trickery used to squeeze in more data is going to increase read and write times. SMR drives are already a big step backwards.
+1 Hopefully HAMR doesn't have the hokiness of either SMR or He-filled drives.
 
Certainly, but next generation SSDs are getting cheaper while next gen HDDs are getting more expensive, so the gap is closing faster than before.

Plus it seems SSDs will outclass HDDs in storage capacity too. 16TB SSDs could come as early as next year while hdds will top at 10TB. HAMR should boost to 100TB but by then SSDs should be past 500TB

HDD is still considerably cheaper at larger capacities today. Until the price differential goes to near zero HDDs aren't going away.
 
Certainly, but next generation SSDs are getting cheaper while next gen HDDs are getting more expensive, so the gap is closing faster than before.

Plus it seems SSDs will outclass HDDs in storage capacity too. 16TB SSDs could come as early as next year while hdds will top at 10TB. HAMR should boost to 100TB but by then SSDs should be past 500TB

Facepalm. No, just no. NAND doesn't scale forever.
 
Not with NAND Flash they aren't. NAND Flash isn't even long-term storage, with data degradation with TLC Flash starting after just a few months. There's also no way at this point to make NAND Flash cheap enough to directly compete with HDDs. The latter will remain 5-10x cheaper per GB for the foreseeable future.

XPoint storage/memory is the only real competition on the horizon for HDDs, and I don't see that one becoming a major threat for the first couple of years.

240GB SSDs have been tested and shown to write over 2 PETABYTES of information before failing. Some reaching even 3 Petabytes.

The rate of wear on these are not nearly as bad as you're making them out to be.
 
Facepalm. No, just no. NAND doesn't scale forever.

It does, but with cost.

Like adding more platters to increase the capacity of a hard drive, you can add more flash to a single board. There's no reason someone can't make a single 'drive' out of multiple SATA SSDs.
 
Watching hard drives grow over the years has been amazing. I remember my first 100MB drive, I thought I would never fill it full. Then, I remember my first 1GB drive, amazed at what seemed like an bottomless pit of storage. Now, I'm running 2TB and need more than ever.

This tech seems neat and all, but, I wonder what can be done if they move away from the standard, and re-invent the hard drive. There has to be something better out there than spinning metal disks!

Agreed - I go back far enough where my first computer was a Mac 512ke that depended solely on floppy drives. I remember looking in the accessories catalogue that came with it, and one of the things you could buy was this massive (in physical size, not capacity) hard drive that gave you a whopping 20 megabytes of storage..... looks like it was this thing: https://en.wikipedia.org/wiki/Hard_Disk_20 and cost $1495!
 
These "future of tech" threads are always hilarious. You can always tell who works in enterprise and who just does this as a hobby.
 
240GB SSDs have been tested and shown to write over 2 PETABYTES of information before failing. Some reaching even 3 Petabytes.

The rate of wear on these are not nearly as bad as you're making them out to be.

Tested with a small pool, once, i wish people would stop assuming every single SSD can handle that work load.
 
240GB SSDs have been tested and shown to write over 2 PETABYTES of information before failing. Some reaching even 3 Petabytes.

The rate of wear on these are not nearly as bad as you're making them out to be.

I didn't mention wear at all. What I am talking about is how long data remains on a NAND Flash chip before it starts to degrade, i.e. when enough of the charge in a cell has been lost that it cannot be read any more. For current NAND Flash (MLC & TLC), this process is so rapid that one cannot write data to an SSD, put it aside (unpowered) for two years and still expect to be able to read the data on it after that time.

That in my eyes at least disqualifies NAND Flash as 'storage'. Short-term, sure, but nothing you'd want to use for off-line storage for any extended period of time.
 
These "future of tech" threads are always hilarious. You can always tell who works in enterprise and who just does this as a hobby.

Yeah...and 20 years ago 4.3GB 10K SCSI HDDs were $600 each. SSDs have blown way past that cost to capacity ratio and continues to get larger for cheaper. There's no way in hell that NAND fabs are going to yield the billions of dollars they've invested in solid state storage tech to aging (and dying) platter tech.
 
These "future of tech" threads are always hilarious. You can always tell who works in enterprise and who just does this as a hobby.

Yup, or a tech-savvy engineer-type. Magnetic storage will be relevant for a lot of applications for a while coming.
 
Sure spinning rust is aging and stuck at 10TB for now won't help, but the cost is always lower, not everyone needs SSD's gains, and when your talking massive storage arrays, SSD's are still far too cost prohibitive for many companys.

Think people who order from Dell, HP, Equalogic, Nimble et cetera, not prices we buy from stores...

A dell SSD has alike a 500% markup... and companies pay that so SSD vs Mechanical is still to expensive for many companies.
 
Facepalm. No, just no. NAND doesn't scale forever.

While NAND doesn't scale infinitely, it certainly has significantly higher density when we talk about physical constrain. The problem with HDD is that you are physically imprinting on a disk surface, and these imprints are significantly larger than a NAND cell. In addition to that, it requires all the hardware like motor, disk platter, read/write head to work. SSD requires none of that. All of the cells are fabricated into small NAND chips, and the only supplementary hardware it needs is a controller chip.

So when it comes to physical constrain, there's just no comparison. NAND is currently limited by the production of NAND memory, and as the process improves, cost will come down while capacity will improve. Plenty of space to cram more NAND chips into a SSD, we just need the cost to come down. While HDD is running out of space, and any attempt to cram more data will push it's cost up.

The only advantage HDD has is cost. However it's hitting it's physical limits, and stuff like HAMR will only add to the cost. And as we see here, they are mighty difficult to do as well. It's really just trying to push the boundaries of a tech that is outdated.

Of course HDD is still the better choice in terms of mass storage at the moment, but that doesn't mean investing money in trying to prolong it's ageing tech is a good investment. That is the point here that I think some are missing. HDD will be the choice until SSD catch up in terms of cost, and then it's time to move on to SSD. No point trying to prolong a tech that is already at it's limit when there are more efficient method out there.
 
It's all about cost efficiency. The only places where SSDs will replace HDDs in the foreseeable future is where the added performance benefit of SSDs outweighs the cost deficit. At least in Enterprise.

At home I can't afford to put everything on SSDs, for putting just my games on SSDs I'd need an 1TB drive at least to comfortable, but that costs around $350 here.

Compare that cost to the average wage of $500/month, and you can clearly see that SSDs as a replacement for spinning drives are just not affordable to most where I live.

Even if the cost difference would be smaller. Why would I choose to buy SSDs for storing images/videos?
 
While HDD is running out of space, and any attempt to cram more data will push it's cost up.

The only advantage HDD has is cost. However it's hitting it's physical limits, and stuff like HAMR will only add to the cost. And as we see here, they are mighty difficult to do as well. It's really just trying to push the boundaries of a tech that is outdated.

Of course HDD is still the better choice in terms of mass storage at the moment, but that doesn't mean investing money in trying to prolong it's ageing tech is a good investment. That is the point here that I think some are missing. HDD will be the choice until SSD catch up in terms of cost, and then it's time to move on to SSD. No point trying to prolong a tech that is already at it's limit when there are more efficient method out there.

Ummmmmm...... HDD's aren't at their limit (see: new tech), and SSD's have a long ways to go in terms of $/gb. Nor is there infinite scaling of silicon.
 
[..]

So when it comes to physical constrain, there's just no comparison. NAND is currently limited by the production of NAND memory, and as the process improves, cost will come down while capacity will improve. Plenty of space to cram more NAND chips into a SSD, we just need the cost to come down. While HDD is running out of space, and any attempt to cram more data will push it's cost up.

[..].

It doesn't appear that you understand the physics behind magnetic storage and NAND Flash. In theory, one only needs to orient singular ferromagnetic molecule in order to store and read out a bit. Flash is far more complex, requiring an elaborate structure to contain a sufficiently large number of electrons in order to record a bit and maintain this containment for as long as possible.

Sadly, containing electrons as a charge is always a losing proposition, as anyone who is familiar with battery and capacitor technology knows. This means that Flash will always be a larger, clumsier and only short-term form of storage than magnetic storage.

Magnetic storage has a lot in common with the resistive RAM (ReRAM) technology used for Intel's XPoint storage/RAM technology, in that it changes a physical property of the underlying material. Such changes tend to be very stable. Magnetic RAM (MRAM) is also used in small quantities at this point.

Meanwhile NOR Flash has been all but abandoned at this point. NAND Flash will be next, as the reality of its complexities compared to the many alternatives catch up with it.
 
My point is this: Spinning disks are, in my opinion, obsolete.

You are, of course, entitled to your opinion, however wrong the assumptions it is based on may be.

Anybody who remembers tape drives should appreciate what I'm saying.

You say that as if tape is dead.
Tape libraries by StorageTek (Oracle) and IBM have a strong presence in the data center. Tapes are by far the most cost effective way for long term data storage. Tape is still so popular that a newer iterations of the LTO standard are raising the capacity per tape.

At my previous place of employment we had just replaced a StorageTek SL8500 with an IBM TS3500. Between 3592 and LTO tapes we had just over 2,600 tapes in production.

At my current place of employment we will be investing in a TS3310 because there's simply no cheaper way to deal with regulatory data retention requirements between 7 and 21 years (depending on data) than to put it on tape and ship it off-site.

Tape generates overhead as tapes have to be periodically tested, no doubt about that. Still, that overhead alone is still less expensive than spinning disk, even if powered down.

What we have here is a difference in from where people are looking. Some of the readers/posters are simply not aware how different technologies continue to be use in certain enterprise environments.

Tapes are decades old and are still around because it's solid mature magnetic technology that serves a specific purpose (long term storage of large amounts of static data). 3.5" or 5.25" floppies are no longer around because their purpose (small amounts of portable data) has been killed by optical storage.

Just because you don't use a particular technology doesn't mean that the market doesn't.
 
As scojer and zadillo aptly pointed out, I too remember the highlights of my computer experiences. The 1541 floppy drive that replaced my tape drive for my Commodore 64, My Atari 520 ST. Amiga 500, etc. My first HDD was 220 MB. I was amazed at how much nudie pics I could fit on it.

My point is this: Spinning disks are, in my opinion, obsolete. SSD's aren't a perfect replacement. Yet. Maybe they will get a lot better over time. Or, maybe, and hopefully, some engineer is out there working on a technology that is so far outside the box that will revolutionize our current two paradigms of data storage.

Anybody who remembers tape drives should appreciate what I'm saying.

I remember tape drives, the problem is you're only thinking of workstation applications of storage. Just because you haven't seen a tape drive attached to a desktop or simple server in 20 years, does not mean that it's gone or even obsolete at this point. HDD's being obsolete any time soon is laughable. You're entitled to your opinion, but in this instance your opinion is irrelevant due to being well... incorrect.

This isn't some situation like a random 40 year old military facility still using floppy disks for something either. Desktop storage needs do not drive the storage market or product development. Do you really think for a moment that hard drive manufacturers are targeting desktop users, even people with large media storage needs, as buyers of 8TB HDDs? Come on now. They aren't selling 8TB SAS drives for $600 to gamers, soccer moms browsing the internet, or even people with massive movie collections sitting on a server at home. Heck, even the $260 Seagate 8TB drive isn't getting bought up by those people. You aren't going to see NAND based SSD's catch up to the storage density, reliability, and cost of HDDs any time soon, period.
 
Ummmmmm...... HDD's aren't at their limit (see: new tech), and SSD's have a long ways to go in terms of $/gb. Nor is there infinite scaling of silicon.

That was the point I was making in my post. HDD could no longer scale without adding new tech, and these new tech will increase the cost of HDD. And as I mentioned too, while semiconductor memory cells do not scale infinitely, they already are ahead in terms of density. The only reason we do not see more capacity at the moment is due to cost. But as far as physical constrain is concern, you can already have more capacity than anything HDD can achieve at the moment. Samsung has already announced a 16TB drive.

And here Seagate is still no where near ready to introduce their new tech.

It doesn't appear that you understand the physics behind magnetic storage and NAND Flash. In theory, one only needs to orient singular ferromagnetic molecule in order to store and read out a bit. Flash is far more complex, requiring an elaborate structure to contain a sufficiently large number of electrons in order to record a bit and maintain this containment for as long as possible.

Sadly, containing electrons as a charge is always a losing proposition, as anyone who is familiar with battery and capacitor technology knows. This means that Flash will always be a larger, clumsier and only short-term form of storage than magnetic storage.

Magnetic storage has a lot in common with the resistive RAM (ReRAM) technology used for Intel's XPoint storage/RAM technology, in that it changes a physical property of the underlying material. Such changes tend to be very stable. Magnetic RAM (MRAM) is also used in small quantities at this point.

Meanwhile NOR Flash has been all but abandoned at this point. NAND Flash will be next, as the reality of its complexities compared to the many alternatives catch up with it.
In theory yeah, but what good is that when the coarse hardware used in HDD can never work with such small magnetic imprint. As we see at the moment, they are having to introduce new method just to further increase it's density, and AFAIK, current method involve stuff like shingled magnetic recording (rather than outright reducing the imprint size).

I agree that XPoint may be the better storage tech, but it's still seminconductor storage. They do not suffer from the same physical constrain that mechanical spinning disk do.

As for long term storage, don't they use magnetic tapes anyway. SSD is certainly not the solution for long term archiving, but I don't think HDD are used for that either.
 
[..]

In theory yeah, but what good is that when the coarse hardware used in HDD can never work with such small magnetic imprint. As we see at the moment, they are having to introduce new method just to further increase it's density, and AFAIK, current method involve stuff like shingled magnetic recording (rather than outright reducing the imprint size).
HAMR is a technique to allow a HDD to use much smaller magnetic areas to store bits using what amounts to basically a small change to the write head (laser for heating). The heating of a small area makes it far easier to flip its polarity than at lower temperatures. The only point here is to make it reliable enough in production, not to make sure it works. HAMR is not new technology, it's just old tech used on a much smaller scale.

You are right that shingled recording is a stop-gap measure, but it is still very useful for archive drives, as there access time is irrelevant.
I agree that XPoint may be the better storage tech, but it's still seminconductor storage. They do not suffer from the same physical constrain that mechanical spinning disk do.
That's why I mentioned MRAM, which is essentially HDD technology (magnetic storage), but using semi-conductor technology. For all we know this is where storage will head to in a few years time.
As for long term storage, don't they use magnetic tapes anyway. SSD is certainly not the solution for long term archiving, but I don't think HDD are used for that either.
For very large data sets tape makes sense, yes. Otherwise HDDs are a very attractive option for archiving purposes, with as mentioned earlier shingled recording HDDs being commonly used for that goal.

As mentioned before, the main issue with Flash is that cannot do anything but short-term storage, because the electrons which define bit levels leak away within a few months to a few years at most. Archival means being able to store something for 20+ years and have it work without issues. Tape and HDDs can do this. Heck, common floppy disks have done this for ages, as many here can agree to from their own experiences.
 
That was the point I was making in my post. HDD could no longer scale without adding new tech, and these new tech will increase the cost of HDD. And as I mentioned too, while semiconductor memory cells do not scale infinitely, they already are ahead in terms of density. The only reason we do not see more capacity at the moment is due to cost. But as far as physical constrain is concern, you can already have more capacity than anything HDD can achieve at the moment. Samsung has already announced a 16TB drive.

And here Seagate is still no where near ready to introduce their new tech.

How is flash evolving without new tech? And, should I understand correctly, the new HDD tech will help HDD's push $/gb further down, which in addition to what Elledan is saying about archival needs, would be where HDD's are going to remain most relevant.

I'm honestly not sure how much lower $/gb is going to go for flash either, as the technologies in place have rapidly caught up with SOTA processing techniques. 3D is hard, and isn't likely going to scale to super-tall stacks, either. SSD may be relatively new tech to the market, but it's piggybacking off the advancements of the rest of the silicon fabrication tech, so trends we're seeing in other silicon tech are very real for flash as well (slowing down quite a bit). Magnetic HDD's have been with us a long time, and have had a chance to shrink time and time again, and it'd be wrong to assume that kind of density scaling is going to happen with flash.
 
Back
Top