News Seagate’s Mozaic 3+ HAMR platform targets 30TB HDDs and beyond

[Admin]

Seagate unveils HAMR HDD platform for 30TB+ HDDs with new media, new heads, and a new controller.

Seagate’s Mozaic 3+ HAMR platform targets 30TB HDDs and beyond : Read more

 

[spongiemaster]

Seagate plans to ship HAMR HDDs for cloud customers in Q1 2023.

Back to the future going on here.

 

[gg83]

Is it even worth investigating in hdd tech anymore? These new mosaic + tech sounds very complex and expensive.

 

[JarredWaltonGPU]

Is it even worth investigating in hdd tech anymore? These new mosaic + tech sounds very complex and expensive.

Basically everything over ~10TB tends to target data centers and cloud usage these days. Consumers can still buy the drives, but unless you want a bunch of capacity, a single modest SSD (2TB to 4TB) is a better option. But cloud storage stuff can easily scale to petabytes and exabytes of storage usage.

 

[PBme]

Is it even worth investigating in hdd tech anymore? These new mosaic + tech sounds very complex and expensive.

Do you think current hard drives seem simple and was cheap to develop? Called technology, always complex and always expensive when creating something new. But even then, this end product, even for early adaptors, is likely a tiny fraction of the cost of what that amount of space in SSD's would be.

 

[thestryker]

While SSDs can scale up storage density wise much further than HDDs could ever hope to HDDs are much cheaper to buy and operate. The majority of HDD technology development has been to increase density which will generally pay for itself over time. There's zero sign that at scale HDDs are in any trouble it's the consumer end of things that has been shrinking. Most of the HDD retail business seems to be external drives or drives for NAS purposes.

 

[LuxZg]

HDD 20TB = 360€
vs
SSD 8TB = 580€

I'd say there's still plenty of space for development and marketing of HDDs.

Like others said, consumer space is now mostly SSDs, except external drives and NAS drives. That's for people that do backups, or store a lot of videos locally (be it movie libraries or personal video archives). Same with some professional markets (audio/photo/video creators and editors).

But cloud providers is where the real need is. As people tend to store data in the cloud, and consume data from the cloud, a lot of those personal photo/video libraries have migrated to the cloud, and a lot of movie libraries as well (in the form of services like Netflix and YouTube). And these need tons of store, and tons of redundant, backup and archival storage. These systems are all tiered, so most read/write requests will hit some kind of solid state fir responsiveness. But after that, a lot of it is HDDs. I found information that in 2020 global storage production was split 5:1 (HDD:SSD), and by now it is probably closer to 4:1. But that still makes 80% of new data systems HDDs. Considering that consumer market is 90% SSDs, enterprise and cloud segment is probably still very much 5:1.

In my company we are almost completely SSD now, both client and server. But all backup and archival is still HDDs, and those few systems (backup arrays on 2 sites, and video surveillance on all sites) despite only being like <5% in amount of devices and CPUs, is about 5x the amount of HDDs vs all client+server SSDs on thousands of devices in the rest of company. Because those systems are in hundreds of terabytes, while production is easily done in <100 TB of solid state store. But keeping multiple copies for archival and backup, revisions, and reasons provisioned by law, makes those 100TB insufficient even for a single weekly backup cycle.

I'd say cloud providers are the same or worse, with same data replicated around the globe multiple times. and all that is on SSDs, while only a single copy is on some hot storage closest to your last access location or whatever.

 

[yankeeDDL]

Basically everything over ~10TB tends to target data centers and cloud usage these days. Consumers can still buy the drives, but unless you want a bunch of capacity, a single modest SSD (2TB to 4TB) is a better option. But cloud storage stuff can easily scale to petabytes and exabytes of storage usage.

Nobody uses backup or storage at home?
I easily have 15TB of photos accumulated over 20 years, and with increased resolutions they keep piling up.
I, for one, would love to see huge capacity HDD for long-term storage. I currently have a NAS (5 years old) with 4x 8TB HDD.

 

[das_stig]

Tell my Plex media server big HDD's are not for the home and it will laugh at you, until big SSDs as cheap as HDDs appear and that's a BIG WHEN !

 

[Tom Sunday]

Is it even worth investigating in hdd tech anymore? These new mosaic + tech sounds very complex and expensive.

My thinking says that when an 8TB SSD is available for $200, then the days of HDDs are clearly numbered, at least in the end-user segment. I can't judge what it looks like in data centers. Costs/TB are just one factor among many. (size, power consumption, waste heat, reliability, etc) However, as soon as HDDs disappear into the niche, their costs are likely to rise and their further development is questionable. The crucial question, however, is whether SSDs will ever come close enough to HDDs in price. NAND prices are currently near a cyclical low again. This making SSD prices currently seem cheaper because manufacturers are selling at a loss.

 

[gg83]

Do you think current hard drives seem simple and was cheap to develop? Called technology, always complex and always expensive when creating something new. But even then, this end product, even for early adaptors, is likely a tiny fraction of the cost of what that amount of space in SSD's would be.

That didn't answer my question at all. I guess I could have been more specific.

 

[gg83]

My thinking says that when an 8TB SSD is available for $200, then the days of HDDs are clearly numbered, at least in the end-user segment. I can't judge what it looks like in data centers. Costs/TB are just one factor among many. (size, power consumption, waste heat, reliability, etc) However, as soon as HDDs disappear into the niche, their costs are likely to rise and their further development is questionable. The crucial question, however, is whether SSDs will ever come close enough to HDDs in price. NAND prices are currently near a cyclical low again. This making SSD prices currently seem cheaper because manufacturers are selling at a loss.

Ah, thank you. That's the perspective I wasn't considering.

 

[Geef]

A very good reason to get a HDD for storage: STEAM.

Play your game on your main drive and when you finish it but don't want to uninstall, you tell Steam to relocate the game to the HDD. This way if your download speed isn't amazing, you'll be able to move the game back to main and play it again quickly.

 

[Uncle-T]

What's the data transfer rate?

What would be the use case for these? Backup/archive storage? CPU/GPU-bound compute jobs where storage access is infrequent?

I certainly can appreciate wanting huge capacity in the form factor of a single drive, but I would wonder if that form factor would impose limits on transfer rates that wouldn't exist at lower data densities (due to its inherent limits on the number of platters, etc.). While there are certainly many workloads that need access to huge amounts of data, access speed is a major issue for many (but not all) of those workloads.

 

[JarredWaltonGPU]

What's the data transfer rate?

I don't know that any information on this is available. SATA still tops out at around 500~550 MB/s, with relatively terrible random access times. Higher areal density can potentially boost sequential read/write speeds, but HAMR, EAMR, and other similar tech may slow down write speeds. The fastest HDDs these days do around 300 MB/s, at the outside tracks on the disks.

What would be the use case for these? Backup/archive storage? CPU/GPU-bound compute jobs where storage access is infrequent?

Backups and situations where speed isn't as important as capacity, yes. Most CPU/GPU workloads that need a lot of data either have to fit in RAM, or else they need SSDs to maintain anything approaching useful speeds. And if each drive uses ~7W of power, doubling capacity while keeping the power roughly the same means you can use half as many drives... or get twice the capacity.

 

[mdd1963]

With warnings of lower IOPS, they've all but stated the drives are/will be slower; will have to see what sort of sustained write speeds can be achieved during RAID rebuilds....; as rebuilding /expanding a RAID5 or 6 with even 14 TB drives can already take 4-5 days in NAS units, no one will be eager to see the 'only 10x longer times required for RAID rebuilds ' debacle of SMR repeated.