Slashdot Mirror
Seagate Reveals 'World's Largest' 60TB SSD (zdnet.com)
An anonymous reader writes: While Samsung has the world's largest commercially available SSD coming in at 15.36TB, Seagate officially has the world's largest SSD for the enterprise. ZDNet reports: "[While Samsung's PM1633a has a 2.5-inch form factor,] Seagate's 60TB Serial Attached SCSI (SAS) SSD on the other hand opts for the familiar HDD 3.5-inch form factor. The company says that its drive has "twice the density and four times the capacity" of Samsung's PM1633a, and is capable of holding up to 400 million photos or 12,000 movies. Seagate thinks the 3.5-inch form factor will be useful for managing changing storage requirements in data centers since it removes the need to support separate form factors for hot and cold data. The company says it could also scale up capacity to 100TB in the same form factor. Seagate says the 60TB SSD is currently only a 'demonstration technology' though it could release the product commercially as early as next year. It hasn't revealed the price of the unit but says it will offer 'the lowest cost per gigabyte for flash available today.'"
hurray more porn!
I'll buy 4 of these for $99 each on black Friday 2021 and put them in a RAID.
Where is the pci-e based one?
More storage density meaning stray neutrons from space (and yes, that's a real 'thing'!!!) could flip a load of bits in one go!
It'll be interesting on how the long term storage/reliability holds up over time. If you don't continually check those CRC's (guessing in idle time) then you'll never know they've been flipped to correct them.
https://en.wikipedia.org/wiki/Soft_error (stuff relevant to this kind of thing)
If you can't afford the baskets, stop collecting the eggs.
That didn't take long. Toshiba announced a 100TB drive (different type) SSD today.
http://www.theregister.co.uk/2...
The end of spindle drives is nigh
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
Why would a tech site such as slashdot ever, EVER, bother with metrics such as "number of photos" or "number of movies". We know how big a Terabyte is. We don't need it spelled out in such mundane, and ambigous terms such as "number of photos".
I can definitely imagine needing the 15TB one in a few years. After being more of a classical literature and music person for most of my life, I've been getting into film. The canon of great films consists of hundreds of titles, at least. In the past you'd have to be lucky to live in a developed country with a well-stocked library, or have a truly massive disposable income to buy all the DVDs yourself. But people today have an incredible opportunity, regardless of their means or location, to educate themselves about this (or any other) art form thanks to torrent communities.
When you're downloading Bluray rips at full quality, where a single film can be 25GB, then storage space starts filling up quickly. One could delete after viewing to save space, but who knows, maybe someday you'll want to watch a particular title again or show it to a friend or loved one, and at that point there might not be any seeders left on the torrent. So, if storage gets cheap enough, then it's worth keeping it all on disk.
Given their reputation, I expect that about a week after I've loaded it full of irreplaceable data (and not-backed-up), it will inexplicably start making clicking noises, and all of my data will be corrupted when read... to die an ignoble death 2 days later with a "pop" and a loud, winding-down whine.
That is why Avago Technologies buyed PEX so they can start pushing pci-e based storage back planes
The cost of being too lazy to back up just went 'way, 'way up.
I've calculated my velocity with such exquisite precision that I have no idea where I am.
Wow, I've been using SSDs since 2009... I've got a 60GB Vertex from that era still running a Linux media server 24/7 - the other of the pair I bought then did die, after 5 years of service. It's the ONLY casualty I've had, and I've got 12 SSDs of various ages and capacities in systems around my home. Compared to platter drives, I've had more success with SSDs.
It's not like they haven't been stress tested by numerous organizations... for all practical purposes, a typical SSD should last even an enthusiast user a decade or more. Probably longer than the life of those platter drive motors.
Consider how ubiquitous they are becoming in the "cloud" space these days. Platter drives are good for one thing these days... archival storage.
I don't know anyone who has or needs anywhere close to this amount if storage.
An enterprise SSD drive is not targeted at individuals, it is targeted at companies. Hence the "enterprise" designation. A single drive that large would potentially replace an entire RAID array, although you're obviously going to trade some performance and reliability, but it's a great place for storing large long-term backup images.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
So, anyone know how many Library's of Congress a beowulf cluster of these would hold for our newly welcomed Solid State Overlords?
You're thinking of individuals, but the very first sentence of the summary says "for the enterprise". Lots of companies have tons of data that they could store on something like this. Since many nerds work for such companies, this seems relevant to their interests.
Lowest price per gigabyte? When you are talking about 60-100TB drives we are now looking at the cost per TB. And we know there are initial setup costs to manufacture but at the end of the day these are still devices made of the cheapest and most plentiful materials on earth and essentially photoreplicated. Sure start it at $10 grand for the suckers like the 2.5 for 6 months but then drop this sucker down to a sane price for the next year for enterprise sales at maybe $400 with a 5 year warranty and to early adopter masses at $200 with a 1 year warranty. By the time that warranty is up make $200 a 3 year warranty and put out a $150 1 year and $100 90 day units. Make 10's of TB the new benchmark fast.
I've already put a very serious dent in my 24TB home array. The problem with a giant array is that although it is very redundant and unlikely to fail your options are pretty limited on where to shift the data to if you want to change something about it.
Mayber higher and higher drive capacities with fast drives will finally start to put some pressure on the companies with 10gbe cores to stop milking ridiculous prices for tech that is ancient. We are very fast approaching the point (in some ways we are already there) where processing is the data bottleneck.
In Soviet Russia, The Solid State stores you.
Is it, really?
I mean, 100TB of spinning rust storage is probably around $3000 or so. 60TB is probably around $2000-ish.
If Seagate and Toshiba are selling SSDs for those prices, then yes, spindle's are dead. But if we're talking about 5 figures or more, then spinning rust has a long life ahead of it.
SSDs are great for plenty of tasks, and the largest ones on the market offer plenty of storage for most users.
However, there are plenty of tasks that demand bulk storage (e.g., media storage, backups, etc) over sheer IOPs or throughput, and demand cheap bulk storage, at that. Spinning rust fulfills that need wonderfully (and there's plenty of demand for it, as well - I'm sure most people have at least a need to have some big bulk storage around to store their media).
If you can afford one, you can afford two.
If only there was some sort of summary or link that would explain why they "hyped" 3.5 over 2.5.
That's a pretty narrow minded view of the world. I'm glad people like you aren't the ones in charge of technical advancement.
Only if it's a buy one, get one free sale.
Competing with hard drives is more than just matching their capacity. You have to come close to their $/TB too. The speed of SSDs make it attractive to replace HDDs for some data sets (boot drive, frequently accessed data, etc.). But if you are storing off-line data that is only accessed once a year, then even 2x the price is way too much. Flash has come down a lot, but it is still something like 8x the price of HDD space. Get back to us when a 1 TB SSD can be bought for $50.
I think capacity per disk is starting to become a bit of a factor, too. For some people (read: large companies), It might be worth spending double the price per TB of storage if it also means you need half as many servers to hold the disks. The savings in hardware costs, rack space, and electricity will make up for some of the additional cost for the hard drives.
I've made the decision a couple times to buy 8 TB hard drives, even if they weren't the best price per TB, because it allows for the highest possible total storage before we would have to spend more on another chassis.
Oh great, now you have a 60TB drive that will fail taking everything with it and because of the size making backups very costly to boot.
Why does it cost more to backup 60 TB of data from a single SSD than it does to backup 60 TB of data from a dozen HDDs?
HDD drives can't fail as a business model because they complement the SSD's weaknesses.
SSD is dependent on firmware and susceptible to software based complete failure, an HDD ain't.
SSD is more likely to become useless after an electrical based failure and in more case than not, the data will be wiped out. An HDD ain't, so long as platters are intact you can retrieve with special equipment.
HDD on the other hand is susceptible to vibration damage, while an SSD ain't.
Precisely because HDDs are slower than SSDs, minimizing any malicious writing can be achieved. While HDDs will always be the main solution in servers while SSDs are at best used as cache, because their electrical vulnerability and their firmware crapup, and lower chance of retreival, make them too high risk. And it ain't ever going to change because of the nature of how the SSD stores information, which is completely different from HDDs.
Saying SSDs will replace HDDs is like saying touchscreen will replace keyboard. Never gonna happen unless under pretense of high delusion.
I know...I could not help myself.
4wdloop
You do realise that SSD reliability increases linearly with capacity, right? Barring manufacturing/design defects (which can affect all products equally), this thing will be far more reliable than any HDD has ever been. A typical life span for TLC SSDs (the very worst kind in terms of reliability) is 2.25kB of write per B of storage. That means that the 60TB model will survive about 135PB of writes before it starts to fail, and the 100TB one will do about 225PB of writing.
So yeh, the 60TB model will sustain writing continuously at 550MB/s for 8 years before it starts to suffer. The 100TB one will sustain it for 13 years.
At more sensible average write rates for a consumer (about 10GB per day), the 60TB one will last about 37,000 years, or the 100TB one will do 62,000 years.
I have no idea why people are still worried about SSDs as being inherently less reliable than HDDs - they're not. They are in fact, much much much much more reliable.
The end of spindle drives is nigh
Perhaps if size and price was related like with HDDs... checking my local pricewatch the cheapest $/GB is a 480GB drive leading by a hair over similar 240GB and 960GB models. Above that 2TB/4TB models actually cost marginally more/GB, probably because of less volume. When you can put 1TB in an M.2 format it's obvious you can go a lot bigger with 2.5" or 3.5" disks. Heck, make a 5.25" SSD for the DVD player bay and you'd probably be approaching the petabyte but it would cost hundreds of thousands of dollars too.
Live today, because you never know what tomorrow brings
While HDDs will always be the main solution in servers while SSDs are at best used as cache, because their electrical vulnerability and their firmware crapup, and lower chance of retreival, make them too high risk.
Our servers run RAID 10 SSDs. If we need more storage we buy bigger/more SSDs.
> How does this affect anyone at all? I don't know anyone who has or needs anywhere close to this amount if storage.
Whether you realize it or not, you probably use multi-petabyte storage arrays daily or have some running nearby without even realizing it.
Walk into a large modern casino with thousands of cameras following your every move, there is probably a multi-petabyte SAN with video footage of you.
Do you use Bank of America, Citicorp, etc? You're accessing data that are stored on SANs that are many petabytes in size.
Do you have a VPS with a larger hosting company? Chances are it's on a half-petabyte or larger SAN.
This will be a huge breakthrough - for data centers they will pay for themselves in energy savings pretty quickly, and also will allow for installation of physically-smaller SANs, with buildouts only being needed as storage bandwidth is saturated.
Plus, Windows' install footprint could be approaching a terabyte in a decade. I'm joking but think it is plausible that this may actually come true. >_>
The Christian Right is Neither (Christian nor right). See: Matthew 23, Matthew 25, Ezekiel 16:48-50
"SSD is dependent on firmware and susceptible to software based complete failure, an HDD ain't."
What the fuck kind of crack are you smoking? http://knowledge.seagate.com/a...
"HDD on the other hand is susceptible to vibration damage, while an SSD ain't."
Hi, my name is ultrasonics, a form of very rapid vibration. I'll make your puny glass chips turn into fucking dust at the right frequency.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Stagnated? 10TB drives are available for purchase. It's easier to scale SSD up, you can simply put more chips in. But the cost is still prohibitive and it will stay that way until you get the price down from $1 to 3 cents/GB.
Custom electronics and digital signage for your business: www.evcircuits.com
... and wait 'till it costs 30 Euros in the bargain bin.
We suffer more in our imagination than in reality. - Seneca
Moore's law applies to processors and has to do with transistors and different corollaries get involved such that the amount of work a processor can do doubles every so-and-so years.
That isn't right.
One failure mode, unrelocatable sector errors goes down with capacity. Other failure modes like bad solder bonds, short circuits, etc. go up with number of chips and bonds in the system.
Since these two curves have different gamma and slope, they have to intersect at some point, where increasing capacity decreases reliability.
I would rather have 12x 6TB SSDs than 1x 60TB SSD. Aside from a single SAS channel not being at all fast enough for a 60TB disk, an RAIM array of SSDs is much more survivable, and takes much less time to recover than a complete restore from backup for a single SSD.
The more ideal thing would be an array of M4 mass carrier boards each connected via SAS or PCIe-ext channels to the machine.
It's not just the reliability of the components, but the reliability of the system; if you have to replace the system as a whole (as in a single 60TB SSD) when one of the components fail, it will be less reliable than a system which allows the components to be replaced at a more granular level. The spring to land connections on M2 decrease reliability, but unit replacement more than makes up the difference in a system with active maintenance.
I have no idea why people are still worried about SSDs as being inherently less reliable than HDDs - they're not. They are in fact, much much much much more reliable.
Not correct, not entirely wrong. HDDs have much lower MTBF, but considerably higher MTTF than SSDs. If you put a HDD away in a suitable safe, you can be fairly certain it will be readable in 15 years time. If you do the same thing with a USB stick or SSD, your chances of data recovery are slim to none. The reason is that SSD suffers from charge migration over time, so without refresh has eventual dataloss, whereas HDD magnetic media has hysteresis which holds the charge permanently when stored away from a permutive force such as a dynamic magnetic field (transformers motors and moving magnets). So basically if you build a huge zfs zvol out of SSDs and resilver it monthly, it will have much higher reliability than a HDD array the same size, but if you just want to put your data away offline for safe keeping, SSD is a non-option and your only options are HDD, tape, or permanent (read super-expensive, not CDR, DVD-R or BD-R) optical media.
So what you are telling me is that HDDs don't have firmware? I think we're done here. Okay, bye.
But seriously, I don't know how many of the HDD failures are due to firmware failure. Also SSDs have higher MTBFs than HDDs and thait is a figure I have seen presented.
That's part of the reason I use external USB 3.0 HDDs.
half as many servers to hold the disks with half of the redundancy.
Now this is cool still need mon servers
http://ceph.com/community/500-...
Rack? We don't need no freaking rack. And as for SAN, I have two 3TB external HDDs hooked up to my laptop that serves as my main machine that I bought for around $400 open box at a Best Buy about 2 years ago. They quasi mirror each other for the harder to replace data and have different parts of my Steam collection and other easy to replace data unmirrored. My mom has a refurbished laptop with a 1TB hard drive with a 1 year warranty from the seller. I've been thinking about stashing some files on her internal drive and getting an external 4TB drive.
I've had three SSD drives fail (all OCZ). Two of them suddenly bricked to the point that they could not be seen as SATA devices. The other one, an "enterprise class" OCZ drive started corrupting itself after two weeks of use. I'd call bricking taking everything with it. I was able to recover data from the self-corrupting drive. I still have two OCZ drives. One is used only for swap and the other gets fully backed up weekly with nightly incremental backups.
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
That is what happens when you run faulty firmware and has nothing to do with flash technology.
Buying OCZ is more closely comparable to running the testing version of a filesystem.
As a mere mortal, I often wonder how many TB per square metre they do in a data centre. Any rough ideas? I dunno what to subtract for all the extra components.
A drive large enough to hold all my porn in the palm of my hand!
A 'singular oddity' is an event that cannot be explained and only happens when you are alone.
What about the RAM needs for this?
1GB per TB is the general rule. So a 60TB drive would demand 60GB of RAM. 100TB drive would need 100GB.
Anyone currently running with 100GB of RAM?
I come here for the love
600TB isn't enough for your pr0n folder? Now that's a collection.
I can do some very approximate math for you.
If you have 42U racks that take up 1 m^2 each (they're about 2x4') and use 4U servers with 36 disk bays (which is the last one I got), you get 360 hard disks per m^2. If you use 8 TB disks, you'd get about 2.8 PB per m^2.
Oh great, now you have a 60TB drive that will fail taking everything with it and because of the size making backups very costly to boot.
The lengths some people go just to post some asinine negative shit. People that can afford these things can typically afford redundancy. These things aren't for stashing your pr0n.
You are joking, but I am sure many of us have seen "home grown" business with mixtures of enterprise grade and home/consumer grade tech that makes us want to cry on many occasions. As long as it "still works" they will fight tooth and nail to keep it just the way it is.
I only look human.
My mother is a halfling and my dad is an ogre, so that makes me an Ogreling
Our servers run RAID 10 SSDs.
How is that working for you? I'm hesitant to go that route, especially for DB servers, because of the stories I've read about mirrored SSDs failing close to one another due to the wear being the same on each. Are you seeing any problems like that, or do you swap out drives before there's a problem or anything?
Why yes, my desktop has 64GB of RAM in it, and it's not even a server let alone an enterprise server.
This is clearly an enterprise product.
If you're doing a lot of random database reads, a low-latency disk with decently high bandwidth is exactly what you want. (Patterned reads should end up cached or prefetched into RAM.)
A larger array of smaller-capacity SSDs would be better for an intensive write environment, but if you primarily need random bits of data very quickly then this will be of interest. (A database write generates more IOs than a read, plus writes often take longer to begin with---so having fewer IOPS is more likely to be acceptable for read-heavy loads.)
This product may appeal to enterprises as an economical purchase if they have a specific workload. Or, if they need as much high-performance storage as they can get, then this will deliver it using less rack space and power.
---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
An unbound neutron is unstable, and decays to an electron and proton (ie - a hydrogen atom) with a half-life of about 15 minutes. Unbound protons, on the other hand, are stable, and are just a hydrogen ion. When they hit the atmosphere at relativistic speeds, they unleash a chain of ionization events among air molecules, which then radiate hard gamma rays, which cause more, but less energetic ionization events, which eventually results in X-rays reaching the surface.
Very often. Plenty of your components generate a ton of ultrasonic noise and vibration. Get a spectral analyzer and proper recording equipment and 'listen' to your computer, phone, etc.
Also, the transfer of energy is a rapid vibration. Too strong of one, your silicon fries (ESD death.)
Basic electronics.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Oh, the ever-wrong AC who's too stupid and such a life failure that all they can do is obsess over me like some strung-out groupie. No job, no life, obviously hasn't left mommy's basement (and could probably never offer proof of having done so in any meaningful manner in the first place,) and obviously no life partner.
Such a pathetic existence you have. It's so pathetic that you have to hide yourself while you fail at mocking me.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
What does the amount of RAM have to do with the amount of connected storage?
Look up cache?
http://saveie6.com/
The cost of rotational storage is more than just the drives - they need a chassis, HBA, backplane, cooling, power etc.
Wow. Thanks.
It is a low power SDD, suited for low access / Archiving.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
IF all you are concerned about is storage capacity, you're probably right. But there are other things to consider, such as IOPs, Energy Costs (spinning, heat/cooling etc) and MTBF rates. Actual VALUE is in SSDs, all things being equal. But if all you look at is Cost / TB, you're right ... for now. My suggestion is that is going to change, very rapidly in the next few years.
You are already seeing Drive Denisities exceeding Spindle drives. You're gonna need eight 8TB spindle drives to match one 60TB SSD. More than eight times the power, cooling, not to mention much much lower IOPs. What good is having all that data if it takes forever (computer time) to get it all off the drive.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
I really, really hope you continue to link to that thread, APK. I know I will. Keep it up with the "third-party" posts too, that totally helps your case.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
Yes APK, I read that exchange too. I'm glad you did, even though you chose to retreat and stop responding, because you're a pussy.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
Yeah, you're definitely not APK, obviously some other random person who supports APK. Because everyone totally uses phrases like "in computing" in normal, everyday conversation. Yep, definitely someone completely different and obviously not the little bitch who got called out on his bullshit. There is not a single person here who would possibly look at that post and assume that it is the same bitch that got called out. It is obviously someone completely different, maybe the same completely different person who started posting and defending APK at the exact same time when you, sorry I mean he, decided to stop posting as himself. Yes, clearly he has many supporters who are definitely not him just being a little bitch because he got called out and exposed. He's far too clever to do anything like that, and definitely has the nuts to create an account and let everyone know when he's posting.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black
No, you dumb bitch, that doesn't show me being exposed as anything. What that shows is that you think that suggesting improvements to a piece of software 16 years ago, and getting paid $100 for a forum post 8 years ago, are things that should be on the top of your resume today. What that shows is you hold up gold stars from your teacher and use those to argue that you're a superior programmer. And you do all of this without even knowing anything about my own work. Seriously, what kind of professional argues that they are better at their job than someone that they know nothing about? In short, it shows you for the bullshit artist that you are. Really though, keep posting that link.
"Our two-party system is like a bowl of shit looking at itself in a mirror." - Lewis Black