Seagate Working To Develop a 100TB Hard Drive By 2030 (cnbc.com) 64
Data storage firm Seagate is working to develop a 100-terabyte hard drive by 2030, touting blistering demand from data centers for the 70-year-old technology in the artificial intelligence boom. From a report: BS Teh, Seagate's chief commercial officer, told CNBC that the company is aiming to launch such a drive -- which would have about three times the capacity of the firm's top-of-the-line hard drives -- by 2030. The largest hard disk drive Seagate currently produces is the 36-terabyte Exos M model, which it launched in January.
"You may be thinking, 'Who would need it?'" Teh said, referring to the idea of a 100-terabyte hard drive. "Well, plenty." He added: "I think there's definitely strong demand. This is a key enabler for the industry to be able to deliver the storage capacity that the market needs, because there's no other technology that's able to produce this capacity of storage technology to meet the growth that the market needs."
"You may be thinking, 'Who would need it?'" Teh said, referring to the idea of a 100-terabyte hard drive. "Well, plenty." He added: "I think there's definitely strong demand. This is a key enabler for the industry to be able to deliver the storage capacity that the market needs, because there's no other technology that's able to produce this capacity of storage technology to meet the growth that the market needs."
Re: (Score:2)
8k porn with smells?
Re: (Score:2)
So slap SSD in front of it.
Re: (Score:3)
I'm pretty sure this is all for datacenters.
The use case for individuals is going to be very low.
Datacenters have multiple levels of disk / storage "caching."
Though these drives are slow, they enable larger and larger libraries of data to be online.
Caching solutions can make the intelligent decisions on what data needs to be fetched, based on heuristics.
Re: (Score:2, Informative)
I'm pretty sure this is all for datacenters. The use case for individuals is going to be very low.
That's what they used to say about the 16TB drives I have in my computer right now.
Re: (Score:2, Informative)
That's what they used to say about the 16TB drives I have in my computer right now.
And the 16GB drives of yester-year, and the 16MB drives of yester-yester-year, and ....
Re: (Score:2)
I'm pretty sure this is all for datacenters.
Datacenters and hyperscalers, at least initially, where the need for backup/archive storage (for AWS, think about the S3 Glacier class storage tier) requirements are for a large amount of storage that ends up being essentially write only (never or rarely read). The fewer racks of storage (and power) the better for those use cases.
Re: (Score:2)
if you split files to be on more than one disk (or mirror) then it's less of a problem... I have a 4 data + 1 parity microsoft storage space at home that gets 600+ purely off mechanical drives. if you wanted to you could do many more than that.. it just creates an issue with minimum amount of drives you have to add to expand storage to an online storage array without degrading performance. cloud providers have resources to do something much more elaborate, of course
Re: (Score:2)
Re: (Score:2)
Sure, it's huge, but what if I need the data "right now?"
Then you should use some other solution. No solution works for all use cases. This is one for massive storage that does not need rapid retrieval (see S3 Glacier).
Re: (Score:3)
They buy SSDs. I can also say this in reverse - sure SSDs are fast, but why should I spend more if I am not planning to transfer the data faster than 100mbps or maybe sometimes 1gbps?
Re: (Score:2)
Re: (Score:2)
There is one place where that tier of data is important, and that is backups. For every gig of data sitting on a SSD, there should be at least two others for 3-2-1 backups, and at least 3-4 for 3-2-1-1-0 [1] backups.
Even home users at least need a NAS with a few TB of RAID 1 protected storage, ideally ZFS for checksumming.
Then, there is offline storage. Hard drives are not designed, nor warranted for archival storage, but what alternatives do we have over LTO? Hopefully China will get those optical drive
Two states of a hard disk. (Score:5, Insightful)
It's either new or full.
I could definitely use larger hard disks at work since we have a few NAS enclosures for backups and replication of backups.
Re: (Score:3)
Re: (Score:1)
Re: (Score:2)
Yes double actuators increase the speed of those drives. That does not mean that speed increase does not to help overall transfer speeds to overcome the larger size. It is still an issue. As an analogy that would be like replacing a lawnmower engine to have twice the horse power. That lawnmower moves twice as fast; it still is not going to challenge a F1 car anytime soon. The current problem again is that even if SAS-5 is finalized and adopted by the time of this drive, the max theoretical transfer rate of
Re: (Score:2)
Not all data needs to be read/written that fast though. Sometimes the data slowly accumulates and just mostly sits there, for example in an email server or a backup server.
Yeah, rebuilding the array would take a long time, but most of the time the transfer rate would be good enough.
Re: (Score:2)
The double actuator drives are useful, but they are more for cases of filesystems like MinIO, where you want to feed MinIO as many spindles as possible, formatting the drives with XFS, because MinIO is handling all the parity and error checking on its own layer, as opposed to a filesystem.
Stuffing double actuator drives into a RAID array can work if done right... that's not its intended use case.
Re: (Score:2)
This is why we have moved from RAID-5, to RAID-6, and now I'm seeing triple parity RAID, as well as RAID 1 with three mirrors, because the transfer speeds for HDDs are not going up as much as capacity, so rebuilds wind up taking longer and longer.
In the past, RAID 5 was good enough. If two drives failed, that was just life, and you restored from tape. Now, we don't have any feasible backup mechanism for most users outside of LTO drives (which are not affordable by the average person), and hard drives (whi
Re: (Score:2)
Screw work, I could use this at home. It would be nice to not have to worry about storage all the time.
Re: (Score:2)
so yes it doesn't have to be fast but you need multiple copies in case of failure a NAS will help speed if the network can handl
Re: (Score:3)
Electricity?
Re:"70-year-old technology" (Score:4, Funny)
Electricity?
Magnets!
I'm sure ICP would have something to say on the subject.
Re: (Score:3)
The wheel!
As a kid, I was brought to an IBM announcement presentation for the 1311 removable pack disk drive, which held 2,000,000 6-bit characters in sectored mode, 2,980,000 in "track record" mode. At an even younger age, I was shown a RAMAC drive on an IBM 650. After that visit, we went out for a fine meal of wooly mammoth.
Re: (Score:3)
The wheel!
As a kid, I was brought to an IBM announcement presentation for the 1311 removable pack disk drive, which held 2,000,000 6-bit characters in sectored mode, 2,980,000 in "track record" mode. At an even younger age, I was shown a RAMAC drive on an IBM 650. After that visit, we went out for a fine meal of wooly mammoth.
Reminds me of my first one megabyte RAM upgrade. The salesman said, "You'll never need that much memory." It was as big as the entire computer case, and needed to have individual chips installed all across its surface before it was the full 1 MB. Man, those were different times.
Re: (Score:2)
This "70-year-old technology" is powered by an even older technology!
"70 year old technology" seems like an insult, but it's really the exact opposite. The underlying technology is still used after 68 years because it works and even though most HDD use cases are disappearing, there are still some viable commercial use cases (mainly cold/warm storage). The viability of HDDs comes from its low cost, a cost difference that SSDs still haven't been able to bridge after several decades. Much of this cost difference comes from arial density improvements, which have slowed in the
Great - I can lose even more data in one go (Score:1)
Re:Great - I can lose even more data in one go (Score:4, Funny)
The last four seagate drives I have bought have bricked themselves within three years and with virtually no use. (I back up everything on multiple hard drives and keep them offline in a safe.) If anything their inactive shelf life seems to be going down. I'm seriously thinking of going back to tape.
Three years? Try their SSDs. You can get that number down to a few months.
Re: (Score:2)
The last four seagate drives I have bought have bricked themselves within three years and with virtually no use. (I back up everything on multiple hard drives and keep them offline in a safe.) If anything their inactive shelf life seems to be going down. I'm seriously thinking of going back to tape.
Three years? Try their SSDs. You can get that number down to a few months.
I'd love an affordable 6 or 8 TB SSD that could replace my last bit of spinning rust... OTOH I'd need at least two as the only disks I've had fail on me in the last few years have been SSDs.
Solid state might be better for long term storage as it's mechanical components that really don't like being idle for long periods (talking from experience of hand cranking an old diesel back to life that's been sitting idle for 8 years). Otherwise, yeah, go back to tape as that's what it's designed for.
Re: (Score:3)
Anecdotes are not data. Aggregated data shows Seagate has good models and bad ones. Fortunately other companies exist, like Toshiba who have good models and bad ones, or HGST who also have good models and bad ones. Some people may even go for Western Digital, well known producers of good models and bad ones.
The reality is your experiences do not matter in the grand scheme of data. There's no obvious reliability issue in any single company. If there was, they would be out of business, especially in 2025 wher
Re: (Score:2)
Re: (Score:2)
I buy only enterprise-grade drives and they seem to work. To be fair, I use them in servers running 24/7 so I would rather just spend a bit more to get the better reliability. I do not particularly care about the manufacturers, so when I need to buy more drives, I just buy whatever is the cheapest that is enterprise-grade and not SMR. So far they all seem to work.
100TB is easy, 100TB in 3.5" 1/3-ht form factor, (Score:1)
not so easy.
On a serious note, if you are willing to use a full-height (as tall as early-1980s 5 1/4" floppy drives) form factor with 3x as many platters as you have now, you could get close to 100TB.
If you want to double or triple the speed-after-cache-exhaustion, add more heads per platter. If you want to increase the real-world speed to SSD-or-faster-speed, add lots and lots of can-survive-power-loss cache (SSD, battery-backed RAM, etc.). But all of this adds costs. For most use cases today, an array
Re: (Score:2)
10 10TB drives will be cheaper to buy, but they will not be cheaper to run, and each bay in a server costs money.
Re: (Score:2)
Pft that's nothing, I just bought a 64TB thumb drive off aliexpress and it only cost $4! I could easily fit a couple of them into the form factor of an SSD
Re: (Score:2)
I think Google wanted to make a form factor where it used 2.5", except very tall, like 3+ inches tall, which would allow for 10-20+ platters in the drive.
I wish this form factor has taken off, because 2.5" seems all but dead, with only 6TB drives being the latest and greatest.
There's already a 100TB SSD (Score:2)
And it was already available 4 years ago. Yes, it's not cheap at $40,000. https://www.techradar.com/best... [techradar.com]
So in 2030, will that 100TB (slow) hard drive be really useful, apart for archiving?
Re: (Score:2)
It will be useful for archiving and other stuff that needs lot of space and does not require high speeds. Unless the price of an SSD of the same capacity drops below that of a HDD, there will be uses for hard drives.
Priorities (Score:5, Insightful)
When I can afford enough storage to record my life from start to finish in 8K, it's time to worry more about indicies and search.
There is a point at which more storage starts to make it impossible to find anything.
Re: Priorities (Score:3)
Re: (Score:2)
Being an AGI assigned to watching someone else live so you can index their life for them to relive at their whim... that's a special hell.
Hopefully we don't figure out how to make an actual intelligence until we're no longer likely to do such things with it.
Re: (Score:2)
On the one hand, that wouldn't require an AGI. Something just slightly better than the current Chatbots should be able to do that.
On the other hand, whether it would be unpleasant or delightful depends on what you like. I can envision an AGI that would like that kind of job, and be upset that it was shoved out of it by cheaper, less intelligent AIs.
That I would find the job intolerable doesn't really speak to even what all other people would feel.
Re: (Score:2)
Yes and no. Larger drives means longer access times. Smaller drives in a cluster with the same overall storage capacity will always perform faster.
Re: (Score:2)
Seek and transfer rates aren't what I'm talking about here. I'm talking about being flooded with so much data you don't know where to look for the particular byte you want in the first place.
Re: (Score:2)
When I can afford enough storage to record my life from start to finish in 8K
One interesting thing that has happened as a result of the Vision Pro release is that video companies have upped the quality of VR releases. It's not uncommon to see sub 1h content now be over 100GB.
You're not storing your life from start to finish on 100TB. You may at best get a couple of weeks in at high quality 8K. You'd have better luck with Real Media in 120p.
Re: (Score:2)
For me it's the ever increasing volumes of data that is the issue. Individual files keep getting bigger. RAW images from my camera, 4k video at 200Mb/sec.
I've started archiving my Laserdiscs too, using a Domesday Duplicator for maximum quality. They come in at about 150 to 200GB/disc.
Re: (Score:2)
Has anyone tried git-annex? It promises to help organize files, as well as synchronize them.
A while back I realized it is often cheaper to just buy more storage than to spend time sorting out files and deleting them. Organizing them is harder though.
Meanwhile... (Score:2)
"a distinguished but elderly scientist" was predicting Petabyte sized drives by 3001.
Wowsers (Score:3)
Imagine the cluster size needed to format this thing in DOS, LOL. But seriously, I remember seeing an ad for a full height 9.1GB drive in a magazine sometime around 1995 and remember thinking, "who in the heck would need a drive that big!?" It was something like $5000 too. It's amazing that spinning rust can be pushed so far.
I'll take 8 of them (Score:2)
6 for storage and 2 for parity.
There is no such thing as enough space. Never has been. Never will be.
Re: (Score:3)
6 for storage and 2 for parity.
There is no such thing as enough space. Never has been. Never will be.
I reckon that no one in their right mind will use these with RAID6 (at least not initialy). ZFS, HADOOPFS or gluster are more likely candidates to be put on top of these HDDs.
Of course, some youtuber (cue linus sebastien) may make a video about them in a janky configuration, and enbarras themselves all the way to the bank.
Re: (Score:2)
I'd probably use them in a good RAID-Z3 configuration, where three would be for parity, nine for data, with a ZIL and L2ARC cache.
Or even better, have hardware RAID that can do read patrols (basically check for bit rot and fix it) handle the triple party RAID 6, because a high end RAID controller has a battery backed up DRAM cache, which can greatly increase performance.
I'm sure someone will toss some in a NAS PC, add UnRAID and mention stuff about it... which at least makes for some baseline video entertai
Just imagine (Score:2)
A Beowolf cluster of these drives!
Transfer rates and IOPs will limit their usability (Score:2)
No mention of improvements to the SATA/SAS attachment or protocols to address how you're going to read or write 100TB in any reasonable amount of time. We'll be lucky to see throughput double, and IOP/s are already pretty much plateaued for spinning disks. All these are gonna be good for is data lakes and backup-to-disk.
Re: (Score:2)
No mention of improvements to the SATA/SAS attachment or protocols to address how you're going to read or write 100TB in any reasonable amount of time. We'll be lucky to see throughput double, and IOP/s are already pretty much plateaued for spinning disks. All these are gonna be good for is data lakes and backup-to-disk.
Both troughput AND IOPS for HDDs doubled in 2020, when dual actuator drives finally hit the market. And the areal desity of a 100TB 8~10 pattler 3.5" HDD will give a nice boots to speed.
While SATA topped at 6GBPS, there is 12 and a rumoured 20GBPS SAS. And there are lab tests on NVMe HDDs.
But I agree with you, unless you put a good cache hierachy (read, write, file, block, object, metadata, journal, tiered) in front of these, data lakes and write once read seldom is the application for these.
Dead Rust. (Score:2)
All my dead spinners are Seagate. I relish the time a 100TB member drops; will take a gigasecond to rebuild... Horrible
Never enough (Score:2)
“Data expands to fill the space available for storage” -- Parkinson's Law of Data [archive.org]
Let me rebuild my RAID (Score:2)
Oops one of the drives failed. Let me just slap a new one in and rebuild the RAID. Wait, the estimated rebuild time is greater than the MTBF for the drive ....
But really, I'm thrilled to see large capacity drives. I might replace my 3x14TB drives soon, although the SMART data is still reporting that they are healthy. I'd like to do a 4 drive RAID 10. I'm too cheap to get a box big enough to try out RAID 6, and I really don't need it. For us playing at home, 4 drive NAS appliances are relatively affordable t
64tb ssd (Score:1)
My disk nodes hold about 800 spinning disk per rack. I'd buy a few racks of these 100TB disks today, but by 2030, I expect 256TB or larger SSDs in half the physical space.