Slashdot Mirror
Intel SSD Roadmap Points To 2TB Drives Arriving In 2014
MojoKid writes "A leaked Intel roadmap for solid state storage technology suggests the company is pushing ahead with its plans to introduce new high-end drives based on cutting-edge NAND flash. It's significant for Intel to be adopting 20nm NAND in its highest-end data center products, because of the challenges smaller NAND nodes present in terms of data retention and reliability. Intel introduced 20nm NAND lower in the product stack over a year ago, but apparently has waited till now to bring 20nm to the highest end. Reportedly, next year, Intel will debut three new drive families — the SSD Pro 2500 Series (codenamed Temple Star), the DC P3500 Series (Pleasantdale) and the DC P3700 Series (Fultondale). The Temple Star family uses the M.2 and M.25 form factors, which are meant to replace the older mSATA form factor for ultrabooks and tablets. The M.2 standard allows more space on PCBs for actual NAND storage and can interface with PCIe, SATA, and USB 3.0-attached storage in the same design. The new high-end enterprise drives, meanwhile, will hit 2TB (up from 800GB), ship in 2.5" and add-in card form factors, and offer vastly improved performance. The current DC S3700 series offers 500MBps writes and 460MBps reads. The DC P3700 will increase this to 2800MBps read and 1700MBps writes. The primary difference between the DC P3500 and DC P3700 families appears to be that the P3700 family will use Intel's High Endurance Technology (HET) MLC, while the DC P3500 family sticks with traditional MLC."
How many write limits does this have?
Sounds good but I'll wait a few years.
These won't do much good in my NAS, as I'm usually constrained by the speed of the wireless network. And with the NAS, I don't need enormous drives on my laptop. I'm not quite sure exactly where these would be used, other than in niche systems that need large amounts of local, superfast storage.
Casually, I wonder what the quality of platter drives will be for the next few years.
It seemed SSD forced HDs to just bulk storage role, but if multi TB SSD drives become common, then HDs, who can't keep up it seems, will have less margin. When that happens I imagine the HD makers would have nothing left to improve on and still make a profit. So they'll cut corners and slowly withdraw from the market.
I tried to find out for a large customer how long the current enterprise SSDs live, but Intel declined to comment. Through the grapevine I have heard of people doing complete replacements every 6 months to prevent failures in production environments, after they learned the hard way that these are not as reliable or long-lived as many people think. Especially small-write endurance seems to be pretty bad.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Looks like the first announced 2242 M.2 drive larger than 128GB, but it's still only 180GB. It'd be really nice to be able to put a 256GB drive where a cache drive normally sits, run the OS and programs from there, and keep a spinning rust (spinning glass, now perpendicular recording is standard) drive in the 2.5" space for media storage. Though by the time 256GB 2242 drives come onto the market, 256GB will probably feel overly restrictive anyway.
You're probably not the target customer. If all you need is a good price to store large personal files then you buy HDDs. If you want to reduce random seek time on a database used by thousands of people connected to your website, then SSDs are an option.
and the mac pro will be stuck at 1TB max 256 min for a year or more with the same price for that time as well.
SSDs aren't for mass storage. You're better off with hard drives or tape for that.
SSDs are for blindingly fast performance first, everything else second. Install your OS and applications on a SSD. Keep your movie and music collection on a hard drive.
Even when the costs do come down, SSDs are likely to be priced at 2x HDD prices, just because they can.
The market prices superior products at higher prices, just because it can.
You might not care, and that's fine, but for those of us who do, we went all SSD awhile ago and love it.
My rule for SSD: I will never again have a non-SSD system disk, due to the significant increase in speed, and also noise reduction (and on laptop increased battery life), you get. The multi-TB storage is on a NAS.
You forgot near zero idle power draw, exceedingly low active power draw for mobile applications. For some, that is the killer feature and fast is just "nice to have".
Help stamp out iliturcy.
Ever since my little EeePC spoiled me, I will never use spinning rust again.
Excuse me, but please get off my Pennisetum Clandestinum, eh!
You done goofed!
My rule for SSD: I will never again have a non-SSD system disk, due to the significant increase in speed, and also noise reduction (and on laptop increased battery life), you get. The multi-TB storage is on a NAS.
Damn skippy!
Chas - The one, the only.
THANK GOD!!!
Given that their competition is 15k RPM drives, you'll be waiting a LONGGG time. They aren't meant to replace large capacity low IOPS drives.
Hate on Intel much? Last I checked their SSDs cost more but were at the top of the reliability list. I dislike SSDs because when they fail they do it catastrophically. I have yet to lose any data from from a failing platter drive. Those will at least give you some warning signs.
Only the State obtains its revenue by coercion. - Murray Rothbard
> You're probably not the target customer.
Well DUH. Delcaring that you are not MADE OF MONEY is a very legitimate sort of thing to say in this kind of discussion. Also it doesn't just apply to "mere individuals". Many if not MOST corporations probably feel the same way.
SSD solutions that are far too expensive to be relevant for most individuals or even corporations are nothing new.
A Pirate and a Puritan look the same on a balance sheet.
Give me an SSD within the same power-of-ten size as a hard drive for the same cost and we'll talk.
Seriously. Give me a 1Tb SSD for the cost of the cheapest XTb hard drive and I'll buy it. But if hard drives get to 10Tb in that time, guess what happens? You then have to give me a 10Tb drive for the same price.
Keep dreaming, buddy-boy. We won't "give you" anything like that for a very long time. The main point in moving to SSDs is R/W performance. Just put an SSD (any size) as your system drive and feel the mindblowing speed difference. In a modern computer, the mechanical hard disk drive is a huge bottleneck: processes spend a lot of time spinning thumbs in "I/O wait" state.
As Dr FrankNfurter says in RHPS "I didn't build him for YOU!!!" It's amusing whenever new datacenter/server technology gets posted on /. that half the posts evaluate the proposed product in terms of how affordable/practical/useful it would be to them in their little client desktop or notebook. All of these Intel drives are intended for server (or at least technical workstation ) use, so they need to be evaluated by ROI they give a business doing high-throughput work. If you think they have great stats but are too expensive, maybe you are not the intended market.
SSD solutions that are far too expensive to be relevant for most individuals or even corporations are nothing new.
You can get an mSATA or M2 small ~32-64GB SSD drive (which many motherboards have direct attach slots for now) for about $60. If you use that as your boot / OS system / critical-app drive and get a slow multi-TB spindle HDD drive for your bulk load-and-save storage you'll get huge improvement in your startup/shutdown times and general system operation while still having cheap mass media. Is that far too expensive?
It's not legitimate, it's completely redundant and irrelevant to comment just to say that you can't afford something. What kind of interesting discussion can follow from that?
P3500 = 374TB for 2TB model = 2 days of continuous writing and drive dies = mlc
P3700 = 50 days of continuous writing = slc
while old Samsung 830 routinely did >1PB with 256GB model.
No, you wont write 20GB per day, those are not home use drives, they go into servers and get killed by bcache.
Who logs in to gdm? Not I, said the duck.
I use them to hold the volumes for the VM's that I'm working on. Server 2008 boots in 10-15 seconds. Everything just runs as if it is running natively on really fast HDD's.
These aren't small VM's either. 150-200Gb and using 8-16Gb Ram.
So, CCD's are for mass storage IMHO
I've upgraded a number of customer machines from HDDs to SSDs and the performance boost is profound, no doubt, and Intel is one of the best performers.
But what's kept me from upgrading my own machine is: encryption support.
The use of hardware computed compression in Intel and other Sandforce SSDs is reportedly at odds with software-based OTF (on the fly) encryption options like TrueCrypt because encrypted data is incompressible, so such benefits are lost. It will probably still be faster than an HDD, but not by nearly the same margin, so you're losing a lot of the performance you paid for.
There's also the question of MTBF because software OTF encryption needs to make a gigantic write operation to initialize an encrypted container and then makes more write ops than in non-encrypted systems.
And any files you place on the drive before setting up encryption are liable to be leaked by wear leveling mechanisms that throw and map pieces of files all over the place, so you always need to start from scratch if you want to be sure.
Some SSDs have hardware based encryption, but it's generally just so they can do a secure wipe by generating a new key. Any options available for user-managed keys are extremely rudimentary and not an acceptable solution. The best option I've heard of is the ability to use the BIOS HDD password as the key and then you're at the mercy of your motherboard and whatever leaks or insufficiencies it has (woohoo, 8 whole characters); not to mention, it really wasn't designed for this purpose.
Nope, they're just not there yet.
I notice that flash is currently goign for about 50 cents a GB and disk about 10 cents. Flahs has been falling faster than disk, but disk still falls too.
Sorry, I didn't get that. I read it in the subject line and then in the body and it left me wanting more. Could you repeat that once again for me please? I'd like it once on the subject line, then once on the body please. That way I get to read the beginning of your post twice, because what you have to say is just so damn important it fucking NEEDS to be read twice, dammit. So, once again, can we have you repeat yourself again, once in the subject line and again in the body, please?
My rule for SSD hasn't changed since their invention.
My rule for storage is I don't trust anyone who can't get the suffixes for bits and bytes straight.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Out first departmental computer in the 1970s has a ten megabyte disk for $15K. And it was the size of a washing machine.
My work laptop uses encryption and was upgraded from an HDD to an SSD about a year ago. The performance upgrade was definitely worth it, and everyone else that got the upgrade agrees. I'm not sure which type of encryption it uses, but it is the kind where you have to type in the password before it boots or lets you do anything.
I also upgraded from an HDD to a non-encrypted SSD in one of my home computers and I would say the performance increase was about the same.
The Official Site of 1337 Pwnage
So: for all practical purposes, the magnetic medium of a mechanical hard drive platter does not degrade at all.
Basically, even though it degrades slightly, I can pretend that it doesn't?
This would mean that, ohhh yes, I'm degrade pretender (ooh-ooh).
Also, does it matter how many Platters the drive has?
"Slashdot - News and Chat Sites Deviant". (Click "homepage" link above for details).
Large RAM cache of database's data used to reduce random seek, in this case nobody cares about SSD or HDD is used.
For typical home use, the write limit will allow five years or more of use. For other use cases, it's a deal breaker. Our servers are an example.
We offer a value priced combination hot spare server and backup solution. We'd like to use SSDs with bcache or something similar. We don't because we'd hit the write limit in three to six months. Write limits need to be 20 times higher before SSDs will work in our application.
> who re-writes and entire HD 2x per day?
We would. Actually at that rate you'd expect it to die within 3-4 years. Drives dying is bad, so you need to replace them BEFORE they are likely to wear out. So figure you can write no more than 1/2 of the capacity per day.
For our hot spare server offering, we use raid arrays of 14 3TB drives, yielding 36 TB.
We'd like to use bcache or dm-cache. With a 500 GB SSD, we could write 250 GB / day - less than 1% of the array's capacity.
Maybe your application, but not for others. SSD's fit well into niche areas in the server market, the last server install I did with SSD's was 5 years ago, they're still going strong. On the otherhand, I've had entire SCSI arrays fail in a year and a half. At least hotswapping makes it less of a pain all the way around.
Om, nomnomnom...
> Maybe your application, but not for others
Absolutely. As I mentioned, typical home use is one application where a quality SSD should be fine. I have a DNS server with several MBs of writes per day, so SSD would be fine for that (virtual) machine.
For the hot spares, no way SSD would work.
> the last server install I did with SSD's was 5 years ago, they're still going strong.
I hope you're checking those drives periodically and have some good monitoring. SSDs from five years ago had an expected lifespan of what, about five years in your application?
What you're looking for is called the eDrive standard. It lets the OS interface with the SSD in such a way as to allow Bitlocker and other whole drive encryption methods while using the SSD controller to do the encryption.
Most laptops max out at 16GB or less of RAM, so a pagefile on SSD is the next best thing.
I've got 16GB in my laptop (Thinkpad E530). I turned off pagefiles in Windows (I do at any RAM level I can get away with, Windows doesn't do them right). Performance is much better with paging turned off and 16 GB of RAM.
Learn to love Alaska
...cutting edge SSD, a proven and stable one would be nice, even if it means two or three year old tech, half the transistor density, a quarter the capacity and a quarter the speed.
It's still a "would be nice to have" here, I'm happy with my Momentus XT spinny-hybrid (that came with the laptop). Not sure what tech's in the netbook, it's only 160GB so I'm not expecting it to be a hybrid, but stranger things have happened. Plus even that's still fast enough to capture an HD stream.
Political debates have me rolling my eyes so much I think I got optical whiplash. I should sue. - Foamy The Squirrel
thats just silly.
running a large table through a scan will reliably blow your cache. for most use cases i/o performance
remains limiting
SSDs are also excellent for mobile devices because they don't suffer catastrophic failures if they are moved while operating.
Many SSDs don't rely on compression to store their data. I've never bought one that did, and as far as I am aware, the problems you are describing were solved a long time ago. The laptops at my office (which are SSDs) are all encrypted currently using sophos full disk encryption, and we are soon moving to bitlocker for it instead (not sure exactly why other than sophos in general just sucks).
>running a large table through a scan will reliably blow your cache. for most use cases i/o performance
remains limiting
Depends on how big your database is. We cache our entire database in RAM because it's only a few gigabytes; reads are as fast as they can be, and, in our case, far more common that writes.
Hating on SSD's still.
Here's the problem, this stems from Operating system behavior and basically applications still thinking they're on Magnetic media.
Operating systems (eg Windows, OS X, Linux) do not treat SSD's as SSD's, they still store log files, swap files and other "repeatedly written" data on the SSD instead of somewhere more durable... or even omitting it if the system has sufficient ram. As long as we're still logging things in 128 byte increments (eg web server logs) SSD's are a no-go for anything with significant users without outright disabling all logging and statistical tracking. The operating system needs to be smarter (and since damn near all Cloud computing VM's are still linux 2.6.x) we're making no headway on this. The operating system needs to know the cell size of the SSD and to only write to it when it replaces the entire cell.
Applications also need to be written like they're going to be stored on SSD's, and for the most part we're good. However if it logs to STDOUT/STDERR or to a debug file it needs to buffer. This should be done at the C/OBJC/C++ Runtime level. Likewise if the software needs to keep state (think about games that auto-save, or MMORPG's that needs to integrity check and patch on a daily basis) it needs to keep this state in RAM, not on disk until the application is ready to terminate, or self-terminates, and only at that time flush the disk buffer.
We have a long way to go before SSD's can replace magnetic drives. Ideally SSD's would be made of two layers, a NAND layer and a SRAM layer, where all readwrite happens on the SRAM layer. Or alternatively "partitioning" SSD's into a NAND and a DRAM layer where all /tmp types of partitions are stored on the DRAM layer and written to the NAND on shutdown. These would need super-capacitors to actually complete a write should the power be lost without a clean shutdown.
SSD is down to about 2x-3x the cost of 15k RPM SAS drives. Which is pretty competitive for situations where you needed the speed of 15k RPM SAS and are short-stroking the drives to get even more speed out of spinning rust.
Wolde you bothe eate your cake, and have your cake?
SSDs are still very fast, even with compressed data (see here). Still many times faster than spinning media. And one pass across the drive to encrypt it is totally inconsequential as far as drive endurance.
I think you're a bit confused over how normal filesystem operations are cached on a modern OS (e.g. OS X, Linux, BSDs, Solaris, etc). Even normal log writing (whether you run it through a compression program or not) on a normal non-SSD-aware filesystem is not going to result in tiny little writes to the SSD. The writes will simply wind up in the buffer cache (or equivalent) and get flushed to media when the filesystem syncer comes around every 30-60 seconds or so. The most highly fragmented case in this situation might require the SSD to flush a 128KB a dozen times each 60 seconds which doesn't even remotely wear it out. Only a complete idiot tries to fsync() a log file on each line, so baring that ... it isn't an issue.
I've heard this complaint many times over the years and not one person has EVER provided any factual information as to what and how much and how often they are actually writing to the SSD. Not once.
The amount of data being written is always an issue with a SSD, but if it's being permanently stored it actually isn't the issue you think it is because the equivalent cost of storage for archival data is actually better with a SSD simply due to the SSD in write-once situation lasting forever (maybe rewrite a full drive once every 5 years or so to refresh the cells x ~1000-3000 rewrites). The SSD will easily last 25 years or longer (probably until the firmware itself degrades), whereas a HDD has to be replaced every 3-5 years whether it's off, idle, or doing work. SSDs are great for archiving stuff. They take virtually no energy when idle and can simply be left attached and powered and the only real wear occurs when you write.
For temporarily staged data... this is probably a SSDs one issue. There is a wear limit after all, so constantly rewriting the drive at a high rate will wear it out. But this is also a problem with an easy solution... since such data is usually laid down linearly and processed linearly, HDDs are still useful as a storage medium. Simpler staging of temporary data doesn't even have to use ANY media if the data trivially fits in ram... you just use a tmpfs mount and schedule a job to process the data at reasonable intervals.
In terms of swap, again you appear to be confused. Simply placing swap on a SSD is not going to wear it out. It depends heavily on how much the OS actually pages data in and out. In most consumer/home-system situations the answer will be 'not often' (relative to the SSD's wear limit). In a server situation swap is not written to under normal operating conditions at all unless someone made a major mistake. It's just there to handle DOS attacks and burst situations in order to allow the system to be tuned to utilize all of its resources as fully as possible.
For example, again a 'tmpfs' (memory filesystem) which is backed by swap can actually be VERY write-efficient since the OS isn't going to flush it to its backing media unless the system is actually under memory pressure. If one schedules things such that the system is not normally under memory pressure (which is a typical case for a server installation), then the SSD won't be worn out... but it will be available for those situations that happen every once in a while that really need it.
Oh well, I don't expect much from Slashdot posters anyway. But, honestly, these things should be obvious to people by now.
-Matt
Depends how much you value your 'movie and music' collection. Once you take into account the fact that your HDDs have to be replaced every few years no matter what (even if powered off and sitting on a shelf), whereas your SSDs are as good as their write-wear limit.
Since movie and music collections are essentially archival data, the SSD is actually a pretty good medium for storing them long-term. You just rewrite the SSD once every few years to keep all the flash cells fresh (on top of adding new videos and music) and it will last many times longer than a HDD.
This narrows the cost factor by a lot. Take (just for comparison) a 1TB HDD for around $90 and a 1TB SSD for around $530. To be conservative lets say that's about a 6:1 cost ratio (SSD:HDD). But when you factor-in the SSDs far longer life span, even being conservative you are still talking 3 HDD replacements per SSD replacement (and more realistically it will be 6+ HDD life cycles but I'll ignore that for now).
The cost difference for your long-term archival storage is now 2:1. The SSD is still double the price, but look at what you get for that: Always on (eats very little power so you can keep it plugged in), NO wear when not writing, NO replacement hassle every few years. Throw in the need for multiple live backups and/or RAID and it becomes a no-brainer.
I'm sold. That makes it worth it.
-Matt