Samsung is now shipping a 15TB whopper of an SSD. Farewell, spinning rust Samsung's 2-5-inch 15.36TB SSD is now shipping, with half as much capacity again as the 10TB 3.5-inch disk drive capacity alternative, and taking up less physical space. This relative monster of an SSD first appeared at the Flash Memory Summit in August 2015. The PM1633a drive houses 512 x 256Gb 48-layer V-NAND chips stacked …
Quote "elsewhere on the internets the figure of 5K (in USD) popped up"
So quick calc, that's about 24p (UK) per GB, which is quite good actually (Using £3,553.57 for the calc). Although that likely doesn't take into account VAT in the UK.
Even at 1-to-1 US to GB (which is typical with tech prices here!) at £5,000, it's still around 33p per GB.
For comparison, current retail SSDs (consumer SATA) are around the 17-25p per GB (from places like CCL/ebuyer etc).
Never mind the cost, think of the raw space savings; if the cost per Gb is within 10% of normal SSDs, you can save racks of space (or expand your SSD storage six times over) without any real penalty - for some people, this could be a gamechanger in terms of their operational ability.
IE take a sixteen bay storage array taking up 3U (probably innacurate but I've not researched serious storage for a while); with 2TB SSDs (for performance purposes) you're limited to 32TB storage, which for some applications (VDI, etc) isn't enough. If you are using an entire 42U rack (give or take space for networking, routing, KVM, whatever - call it 36U), you get 384tb.
Now, replace those 2tb SSDs with 15tb SSDs. You can get in 6U what you previously needed more than an entire rack for. Meditate on the power usage, and think where else you could direct that - better networking, more compute to process the data, etc.
I'm loving this; it's what we've been hoping 3D NAND would bring to the table.
Now, lets think about the 'home gamer' - some chump like me who wants to play with virtualiation, biggish DBs, etc. Lets have a consumer version of this, so I can stick it in my NAS and have iSCSI shares that are actually usable to a realistic extent, unlike what you can get from spinning rust at the moment...with double the capacity that's currently feasible with HDDs.
Awsume.
Steven "ALL THE STORAGE" R
"Now, replace those 2tb SSDs with 15tb SSDs. You can get in 6U what you previously needed more than an entire rack for."
12 years ago I purchased a 10TB storage solution. The drives alone were 2/3 of the rack (OK, OK, they were MSA1000s rather than disk drawers, but disk drawers weren't available then.
Now I can buy half as much again in a single 2.5" case with better performance.
"Now I can buy half as much again in a single 2.5" case with better performance."
(5 months later, when I have real pricing and specs)
Half as much again, in a single 2.5" case with 10,000 times the performance (conservatively) and at 7% of the price of the array I purchased 12 years ago.
Oh and it draws 4watts max(ok, 100 watts wrapped in a server), vs 5kW
If you need to ask the price, you probably can't afford it.
The next year is going to be interesting in SSD land.
And don't anyone ask for a SATA version!
SATA needs to be treated as a legacy interface reserved for spinning rust and plastic only, or for handmedown SSDs. not for newer SSDs.
Do you mean ZX Microdrive ? Or when Sinclair was proposing doing Wafer Scale Integration (shipping a whole silicon wafer of chips as one device for storage)?
Or maybe Wafer Scale Integration was what the ZX Microdrive was supposed to be (before it became a loop of video tape)? And you remember more than I do ?
@simpfeld
I'd forgotten about the single wafer storage system, so thanks for reminding me. What I was alluding to was the original Microdrive demonstration where some wags were suggesting that Sinclair had been having typical teething troubles and was using Microdrive casings filled with RAM chips.
But you are missing the point. As larger enterprise storage moves to SSD, the consumer grade devices will fall in price while growing in capacity and reliability. I see this year as the year I convert all my home machines to SSDs. No need to buy a new PC or laptop when you can gain huge improvements for a couple hundred.
Make sure you have a VERY regular back up of any important data you store on them; because unlike spinning rust, you may get NO warning of failure.
I had a 3 week old Enterprise Class SSD drive totally fail at the end of January; system crashed without warning and on reboot the drive was dead, not even recognised as a drive.
Which reminds me, I havent sent the bugger back yet.
Really? I've always found RAM upgrades to be the best. SSDs make applications open a bit quicker, and games load a bit faster between levels. Other than that they're not a big deal.
Enterprise applications, on the other hand, with constant I/O, benefit greatly.
Oh, I guess they're handy if you reboot your computer a lot. I tend to do so once a month a so, so no big deal. Having said that, SSDs are now cheap enough that I have two laptops and two desktop PCs booting from SSD. The biggest benefit for me at least is the fact they make no noise.
Toshiba introduced some stacked flash chips last year using TSVs, so it is possible Samsung is doing this also but I don't really know. They may be stacked traditionally, the main benefit is performance but that's irrelevant when this is slowed down by the SATA interface anyway. It also helps a little with power but that's not a huge concern with flash.
This device exists, it can be made, it has been made.
The price is high because people are willing to pay for it - Samsung has spent gazillions on R&D, fabbing and all the rest that they need to recover - but the actual cost of physically producing these things is I suspect pretty reasonable.
I can't see spinning discs ever beating this on capacity. They can shingle, they can fill it with Helium, but magnetic drives are now officially dead. I'm calling it.
I will not mourn their passing.
http://www.cdrinfo.com/Sections/News/Print.aspx?NewsId=42245
Hard drives can go past 15 TB using shingles, bit-patterned media, hamr, two dimensional magnetic recording, etc. - and maybe several at the same time. But at the rate SSDs are growing, they will never catch up. This 15.36 TB is before generational improvements to V-NAND that will include 100+ layers (thousands?) and shrinking back to 1X nm. They will still have high endurance when they shrink because they will be able to massively over-provision using the layers.
By the time hard drives manage to lumber on to 100 terabytes (in a large form factor than SSDs), there may be post-NAND/Crossbar/XPoint available. Ultimately, holographic storage will blow past what hard drives can store.
http://www.nature.com/news/2009/090124/full/news.2009.54.html
https://en.wikipedia.org/wiki/Areal_density_(computer_storage)#Research
It does look like the writing's on the wall for spinning disk, but it's been that way for a long time. The amount of data being stored is increasing at the rate the price is coming down, but that's the same no matter what medium is used for storage. It's the total cost that matters; not just the cost of the device itself, but all of the other stuff: power, cooling, rack space, reliability etc. And it's a different calculation for different applications; even in the enterprise, few applications actually need the performance offered by SSDs, which is why data centres are not wall-to-wall flash. If there reaches a point where flash is at the same price point overall as rust, then the switchover is inevitable, but it's not going to happen much before.
If you sweat your assets for 3-4 years and are buying now, I reckon in two generations it will be all solid-state. Right now, if you have mixed workloads (as most people do) then hybrid's the way to go, at least in the enterprise.
"Later this year Samsung will ship 480GB, 960GB, 1.92TB, 3.84TB, and 7.68TB versions of the drive. It expects this drive "to rapidly become the overwhelming favourite over hard disks for enterprise storage systems," not differentiating between performance and capacity data storage in its statement."
I assume this means for the user base market yes?
I've been putting off NAS drive upgrades for a few months now waiting for WD 8TB Red's to come out. Think I may just wait that little bit longer for these bad boys.
From Webster:
Definition of half as big/much/good as
—used to say that the size, amount, or quality of one thing is half or nearly half that of another <The bush is half as tall as the tree.> <This dress costs half as much as that one.> <If 100 people were expected and 50 came, only half as many people came as were expected.> <If he can play the guitar half as well as he can sing, he should have quite a career.>
So finally a 5 GB version ???
These drives offer a huge increase in compute and storage density for Analytics applications - but only 32K IOPS - which is probably measured at full Q-Depth - seems low for these high intensity operations. I would otherwise place tow of these on blades in a chassis and deliver 30TB per blade - enabling HP or Cisco Blade systems to stuff 480TB to 240TB into a single chassis. Then it is just a matter of time before this approach faces competition from Intel's effort to marry 3DXpoint onto their CPU's - providing 1-2TB per CPU. I would hate to be in the traditional array business these days.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
