Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Hardware

Cray's New Solid State Storage 382

Sivar writes: "Cray, a well known vendor of extremely fast supercomputing hardware, has introduced a storage system with a 224 GB capacity. The large size seems impressive, but the device can also transfer an unprecedented 80GB(!!) every second. That's more bandwidth than the main memory of most servers, and it's just for storage. For comparison's sake, a typical dual channel DDR motherboard has a bandwidth capacity of barely 4.2GB/sec." Yow.
This discussion has been archived. No new comments can be posted.

Cray's New Solid State Storage

Comments Filter:
  • This is more of a conglomeration of current technology into a pricy solution, not so much a stellar advance.
  • by mESSDan ( 302670 ) on Monday April 08, 2002 @06:51PM (#3306170) Homepage
    C'mon, where are the obligatory Quake3/UT FPS statistics? I want to know if I'm going to get 1,000,045 FPS or 1,000,0053 FPS. Don't they read Tom's Hardware?

    Heh

  • Wow (Score:2, Interesting)

    by dspeyer ( 531333 )
    I'm impressed.

    Of course, it probably won't work on ordinary computers (after all, sticking that onto a SCSI bus would be sort of a waste), but eventually we'll get our hands on this stuff.

    Anybody dare to ask how much it costs?
    • Don't ask.... (Score:3, Interesting)

      by rlwhite ( 219604 )
      "Nobody who has to ask what a yacht costs has any business owning one." -J.P. Morgan

      Why else do you think a company with expensive products like Cray's would avoid posting prices online?
  • Hmmm... (Score:3, Funny)

    by Stinky Boy ( 107316 ) <stinkyboy@@@comcast...net> on Monday April 08, 2002 @06:56PM (#3306219)
    None on EBay yet...

    Well, looks like I'll have to wait a few weeks.
  • Please tell me that cray.com isn't slashdotted. Oh, how the mighty have fallen!
  • /.'ed (Score:4, Funny)

    by BlackSol ( 26036 ) on Monday April 08, 2002 @06:59PM (#3306242)
    anyone else find it hilarious that the site is slashdotted?

    I know its probably hosted by someone else but come on just the idea that we slashdotted a cray is awesome :)
    • I know its probably hosted by someone else but come on just the idea that we slashdotted a cray is awesome :)

      Crays do vector computing. Why the hell would they have a lot of bandwidth? Though, the day we slashdot UUnet or AT&T I'll laugh until I pee myself.
  • by Sivar ( 316343 ) <charlesnburns[.]gmail@com> on Monday April 08, 2002 @06:59PM (#3306244)
    Get one of these, downgrade your system to 8MB RAM, and run everything from swap...

    Watch your system's responsiveness double.
    • why use any RAM at all? That would be an interesting turn of events, wouldn't it? As a matter of fact, that might be an actual Paradigm Shift in computer architecture.

      • Back in the 60's, non-volatile magnetic core memory was used instead of RAM. So the idea of non-volatile memory is actually very old.

        With virtual memory hardware, you can write an operating system that simulates non-volatile main memory, using hard disk as a backing store. What you get is a Persistent Operating System [arizona.edu]. You don't need a file system. Instead, you store data structures in main memory, and they persist forever, surviving reboots.

        Doug Moen.

      • That's the way PalmOS computers work today. There is no difference between your short-term storage and your long-term storage. That's why, with a Palm application, when you enter an event or click a checkbox, you don't have to "save" the results, because it's all in the same memory area.

    • Too many context switches makes your system go dog slow. While the transfer rate is extremely high, the latency of talking to a piece of storage that is perhaps several feet away, at the speed of light, is too high.
    • ummm. . . no (Score:5, Informative)

      by Bastian ( 66383 ) on Monday April 08, 2002 @07:32PM (#3306464)
      The CPU gets stuff from the cache.

      The cache gets stuff from the RAM.

      The RAM gets stuff from the hard drive.

      The solid state machine won't act like faster memory, making cache misses cost less. It will act like a faster hard drive, making page faults cost less. Using this stuff as a substitute for RAM will slow down your computer unless you have it hard-wired into your system's bus in place of RAM.
    • Um, not with any current OS or standard PC hardware. If it treated the SSD as a hard drive, it would insist on transfering everything into those 8MB before operating on it. So you'd beat the hell out of your 8MB of RAM, and you'd be doing it at RAM speeds, not 80GB/sec. In this completely hypothetical setup, it would definitely be faster to have more RAM. No PC could probably connect to this device, let alone exercise it.
  • Anyone find this comment from the release kind of ironic?

    The field-upgradeable SSD system can hold 27 copies of the Human Genome and transfer data at a rate equivalent to 100 Human Genomes per second.

    yum!

  • by MikeyLikesIt! ( 313421 ) on Monday April 08, 2002 @07:03PM (#3306275) Homepage
    The field-upgradeable SSD system can hold 27 copies of the Human Genome and transfer data at a rate equivalent to 100 Human Genomes per second.

    I guess that using standard measurements (GB and GB/sec) just isn't intuitive enough! But why use the humane genome as a reference? Is that REALLY more intuitive to most people? Does anyone (besides geneticists) really understand how much information is in the human genome?

    • that is the absolute worst reference "unit" I've ever seen! Depending on how you define "human genome" it's anywhere between several hundred MB to several hundred GB. Can we just go back to the more reliable and accepted units, like the LOC (Libraries of Congress)?
      • More reliable and acceptable units would have to be in football fields. Of course, I measure my velocity in nano-parsec per micro-fortnights, but that's not a bad unit. I suppose for measuring the number of bits, one would need a good conversion for surface area as in GB per sq in... thus 240 GB could be construed as something like 2 micro-football fields.
    • Is that REALLY more intuitive to most people?

      If this were a consumer device, I'd probably agree with you. But who, besides geneticists, is gonna see one of these anytime soon?
    • They couldn't care less what "most people" think - they're selling to a small, sophisticated set of users which pay to have the performance from the future today. I'm a little surprised they mentioned any scale whatsoever...
    • by Joel Ironstone ( 161342 ) on Monday April 08, 2002 @07:25PM (#3306426)
      I can transmit my whole genome in a few seconds, While the silicone guys find ways of speeding this benchmark up, I'm looking for ways of slowing it down.

    • Ok, if we're going to measure capacity in terms of Human Genomes, I want to know how many Jelly Donut Units per hour it takes to power this thing.

      "From the intensity of the flame we can deduce that this was a particularly delicious donut."
    • Because that's who they're marketing it to - the bioinformatics crowd. Try to forget that there is little-to-no evidence that there's actually a market for all this genome-crunching that's going on.

      Or that there are real questions about the quality of the sequencing that's been done, not to mention the (abysmal) quality of the code being written to analyze the sequences.

      Bioinformatics is the dot-com boom all over again...

      -Mark
      • by jonbrewer ( 11894 ) on Monday April 08, 2002 @09:43PM (#3307110) Homepage
        (to diverge ever so slightly)

        "Bioinformatics is the dot-com boom all over again..."

        I think not.

        There is quite a market for bioinformatics. My employer spends around 5 billion USD a year on pharma R+D. Much of that money is used in traditional "brute-force" type attacks of screening many compounds against many targets.

        There is tremendous potential for savings through bioinformatics, and the evidence is working its way through pharma pipelines as we speak.

        While there may be as much hype around bioinformatics, the field is solving a genuine problem for a mature, well-funded industry, unlike the dot-com book which speclated on products many didn't want with money that didn't exist.
    • But why use the humane genome as a reference? Is that REALLY more intuitive to most people? Does anyone (besides geneticists) really understand how much information is in the human genome?

      Of course most non-biologists don't really understand just how big the human genome is. That's why they're using it as a reference. The genome is actually smaller than many people think (about 3 GB at one base per byte, but trivially compressible to 1/4 that), which means that expressing things in terms of the genome is a good way making your equipment sound more impressive than it actually is.

  • Yeah Yeah... (Score:2, Insightful)

    by Loki_1929 ( 550940 )
    But can it run Windows fast?

  • Overhead time? (Score:2, Insightful)

    by truesaer ( 135079 )
    80GB per second is impressive, but the transfer rate of existing drives is also plenty fast. The problem is that setup time, head seeks, and rotational delay make it slow for most data accesses which are small. This is of course the whole point of ram, caching, etc.

    Probably this is just useful for transfers of very large amounts of data, and is the same as other storage devices except for its large size...

    • Re:Overhead time? (Score:3, Interesting)

      by akula1 ( 463239 )
      Since this is a solid state storage device its performance will be that of a RAM. This is the main reason why solid state storage is so attractive. There will be no read/write heads etc...
      • Since this is a solid state storage device its performance will be that of a RAM. This is the main reason why solid state storage is so attractive. There will be no read/write heads etc...

        Not all solid state stuff has fast writes. Look at FLASH for example, can be designed to read really fast, but in the general case you have to clear blocks before writing, and the clear is slow.

    • Re:Overhead time? (Score:2, Insightful)

      by Dr. Weird ( 566938 )
      Just useful for transfers of very large amounts of data? Gee, is that what a high speed, 224 GB HD is good at? ;-)

      As someone who routinely works with large datasets ( > terrabyte uncompressed) as is typical for physical simulations I would LOVE to have one of these.
    • "Solid State" makes all the
      difference...although I know it is hard nowadays to damage a laptop HDD just by operating it in a car, there were those days a slight knock on a HDD would scratch the disk surface...

      "Solid State" gives you the peace of mind that HDD never will - knowing that you data does not need that wimpy protection from that thin layer of air is good enough - that your data won't be loss all of a sudden if you kick your computer by accident...(although it is hard nowadays, you never know)...

      I know some Solid State storages like CF can only survive so many writes, but that's PREDICTABLE. For a limited time you just KNOW that your data is safe. I can't say that for any HDD, no matter how many G's of shock it can take.
    • The problem is that setup time, head seeks, and rotational delay make it slow for most data accesses which are small.

      The reason why these things exist is because of the mechanical components that make up drives. Solid state storage shouldn't have the same limitation.
  • by geekoid ( 135745 ) <dadinportland AT yahoo DOT com> on Monday April 08, 2002 @07:07PM (#3306311) Homepage Journal
    Library of congress is that...
  • Bah. (Score:2, Informative)

    by dkresge ( 472373 )
    So, is the 80GB/s aggregate access for the (n) processors in the box? It's a Solid State Disk -- In other words, it's memory. And, it's not _that_ fast for a shared memory architecture system. see: STREAM Memory benchmarks [virginia.edu]
    • Wow those stream numbers suddenly look really pathetic. On a 32-way DEC^WCompaq EV7, STREAM hauls down 200,000GB/s, or 300 times faster than the top supercomputer in the list you cite.
  • How much is it, and can I get it as birthday present?
  • check my math? (Score:4, Insightful)

    by doooras ( 543177 ) on Monday April 08, 2002 @07:19PM (#3306387)
    ok, the article says that it can hold 27 human genomes.

    it also says that it can transfer 100 genomes every second.

    224/27 = 8.29ish

    at 80GB/sec, wouldn't that be about 10 per second?

    or am a moron?
  • shhhh (Score:4, Funny)

    by Sebastopol ( 189276 ) on Monday April 08, 2002 @07:21PM (#3306396) Homepage
    i bet its a hell of a lot quieter than our 500 GB array built on 2GB barracuda drives!!! the first few days i sat near it, it drove me batty, by now the tidal sounds of slightly-out-of-synch cylinder searchs ... that placid sequential tapping ... has an almost embryonic quality ... (my mother was a data entry clerk for the city).

    although I like the new metric: # of human genomes per second ... genops? hugenops (hyoo gen ops)? Huguenots?

  • How Much (Score:3, Interesting)

    by meggito ( 516763 ) <npt23@drexel.edu> on Monday April 08, 2002 @07:22PM (#3306407) Homepage
    I'd really like to see the prices on something like this. But then again, they are probably fairly customizable so the price may be flexible. But still, how about giving people a ballpark or 'starting at' kind of thing. Is this 10s of thousands, or 100s, or in the millions?
  • This will only make your computer faster for 3 seconds, then it will be full. Nice idea, but lets talk petabytes!
  • by Bastian ( 66383 ) on Monday April 08, 2002 @07:41PM (#3306526)
    ncludes a 224-gigabyte Solid State Disk (SSD) with a data transfer rate of 80 gigabytes per
    second


    can hold 27 copies of the Human Genome and transfer data at a rate equivalent to 100
    Human Genomes per second


    Ok, so can it hold more data than it can transfer in a second, or can it transfer more data in a second than it can hold? Pick one, boys.
    • Obvious typo. Probably meant to say it could hold 270 copies of the Human Genome. Although I seem to remember the Human Genome being slightly biggerthan the .8GB they are claiming it to be.

      *me looks aside to the 5 640MB cd's labeled 'Human Genome'.
    • tingling ? (Score:3, Funny)

      by Erris ( 531066 )
      Marketroid: Clever observation. Of course it all depends on which Genome we are talking about really. Would that be an African or a European swallow?

      Bastian: I don't knooooowwWWWWWW, AAAHHhhhhhh! Bastian is trown from the clif by an an invisible hand.

      Those struck by lightening and survive fear tingling sensations.

  • I've looked all through the Cray website, and I can't find the online order form. How am I going to get one of these systems FedEx'd to me now?
    • Re:well. damn. (Score:2, Interesting)

      by Jouster ( 144775 )
      Actually, you have to email ussales@cray.com [mailto]. Here's a recent email I received from them regarding prices:
      Date: Fri, 29 Mar 2002 12:51:56 -0600
      From: Cathy Wells
      Subject: Re: Prices
      To: Dan Reif
      Organization: Cray Inc.
      X-Mailer: Mozilla 4.79 [en] (Windows NT 5.0; U)
      X-Cray-VirusStatus: clean

      The Cray SV-1 vector supercomputer is the current production product for vector types of application problems and most legacy Cray scientific and engineering applications. A 32 processor SV-1 with 32 megabytes of shared main memory would have a list price over $3M depending on the configuration and peripherals required.

      The Cray T3E massively parallel supercomputer continues to be the world record holder for the fastest sustained performance on a real world application. These systems can scale to over 2000 processors. An entry level configuration for the Cray T3E might start at approximately $2M with a 1000 processor system listing at over $30M depending on the configuration.

      The MTA-2 will be available 3Q 2001. It will be an all CMOS machine--much simpler to manufacture than the all GaS system at SDSC. The MTA-2 will have between 16 - 256 processors and 64 - 1024 GBytes of shared memory. We plan to build the MTA-2 in 16P cages. A case, or stack as we're calling it, will house 4 cages. The case will be approximately 3 ft wide, 5 ft deep, and 7 ft tall. It will weigh a few thousand pounds.
      Jouster
  • If they can make a storage system that moves 80GB/sec, why are we still putzin' along with DDR RAM that only moves 4.2GB/sec ???

    Methinks were being jipped...

  • by wringles ( 12507 ) on Monday April 08, 2002 @08:51PM (#3306858)
    "System Can Transfer 100 Copies of Human Genome Per Second"

    Big deal, I can transfer O( 1e6 ) half-copies of the human genome in less than five minutes.
  • The announcement was pretty thin on technical details. What exactly is meant by "Solid State Disk." Are there spinning platters? That title implies not, to me. Exactly what technology allows 224G of storage in non-platter form? Is this an actual commercial implementation of the crystal holography gunk and other amazing "future" stuff?

    It sure would be interesting to know if this is a real advance, or just a big disk.

    -me
    • Re:Tech details (Score:3, Informative)

      by tcc ( 140386 )
      >What exactly is meant by "Solid State Disk." Are there spinning platters?

      No moving parts, you can look at this like a "big ram disk" exept it has it's interface like another storage device. Look at this like A compactflash for example (it's not "SSD" but it's a good comparison.

      There are a lot of interfaces (PCI, ATA, SCSI, proprietary (80GB/sec is either a big aggregated pile of raids or something similar) for these "drives" at various price points. The advantage of a SSD drive on a PC is that you have instant access, and it moves the stuff at a lightning speed limited only by your bus. Let's say you run a SSD drive on a Ultra160 interface, what you'd probably see with a disk benchmarking tool is 100nS access time (versus ~10ms for a standard drive) and you could see the real-world numbers of your scsi bus, probably around 140MB/sec (didn't try one on a U160 bus). The application for these babies are numerous: instant access to data on boards that don't handle 100GB of ram to cache everything or you wanting the machine to preload 1 hour at every reboot, bandwidth hungry application (although a raid could do the same here, but I saw some specific application needed both the bandwidth and under 1ms access time needed so..), for heavy swapping of numbers without using a buttload of ram again, etc.. probably some other people could think of something other. Usually when you break a certain amount of GB, the drives becomes cheaper than a motherboard that could handle a load of ram and the ram modules themselves, so it makes more sense if $$ is a factor (but still it's very expensive, we're talking 10K+ easily for a few GB).

      There's also plenty of product on the net (search google), like I said, some are PCI cards that you add to your system with PC100 ram on it, some are IDE/SCSI, etc.. But for home people, you'd be better off with a cheap IDE raid card and a few drives, it's way cheaper :).
  • It seems to me (Score:4, Informative)

    by compupc1 ( 138208 ) on Monday April 08, 2002 @09:09PM (#3306956)
    It seems to me that many people here have no idea what a true supercomputer (or more specifically, a Cray) is or what makes it different from a simple cluster. Here's a few things to think about:

    - Crays do not have monitors. They do not have keyboards, or mice.
    - Crays do not run Windows. Crays do not run Linux. Crays usually run UNICOS, a special *nix designed specifically for Crays.
    - Crays communicate with the outside world through a host terminal, like a SGI workstation, or something similar to that. Crays DON'T HAVE CD-ROM DRIVES!
    - Nobody but those with 8-9 figure incomes get to buy a Cray. They cost MILLIONS, and the higher end ones can cost many many tens of millions.
    - Pretty much the type of people that WOULD buy a Cray would be the government, and very very large corporations. Sorry, guys.
    - Simply connecting 30 PCs together in a cluster will result in a nice, fast supercluster, but it won't come close to a Cray, because Crays are designed from the beginning to be as parallel as possible. Face it: beowulf clusters really can't make the best use of the contained hardware because the hardware wasn't designed to be so distributed.
    - Be impressed with Crays. Be very impressed.
    • Pretty much the type of people that WOULD buy a Cray would be the government, and very very large corporations. Sorry, guys.

      Even then, Cray cannot sell them to many countries because of legal restrictions... :-)
  • by Anonymous Coward
    Read the press release:

    "With their 32-gigabyte central memories..."

    Of course they need a 224 GB "solid state device" ! Every worthwile competitor of theirs can just put 256GB of main memory in their big box.

    It looks to me that Cray can't easily address more than 32GB on their box, so they just use "extended memory" as a disk.

    Buy an IBM / HP / Sun top of the line, stack
    it with 256GB, and you can use 224GB as a file buffer. Or 128GB, or 16 GB, or whatever you do not use for something more important.

    You've been fooled by PR spin on a limitation :-)

    Like windows and 36bit addressing on Xeons...

    • by angio ( 33504 ) on Tuesday April 09, 2002 @01:00AM (#3308060) Homepage
      Not really - consider SGI's servers, for instance. The Origin 3800 can handle 1 TB of RAM -- but it's a CC-NUMA machine, meaning you have to go through an intermediate router (don't think Internet; much faster) to get to the memory. SGI machines have a limit of 8GB per processor "brick", and their bricks interconnect at 1.6 or 1.2GB/s.

      Then consider the SunFire 15K - it's an SMP machine; processors fit on boards that can contain up to 32GB of RAM; after that, you have to go off-board through a switch to get to other memory. Each system board has about 9.6GB/s of offboard memory access speed.

      In short, Cray isn't tooting needlessly - this is impressive bandwidth to the memory. Latency is probably fairly high on it, but for streaming vast quantities of data in and out of local storage, it's probably amazingly nice.
  • But I'm picking up to after work!

    /dies

    I'm not sure what happened, but my coworker just screamed "the price" and died in his cubicle?
  • We had this discussion on slashdot back in the day [slashdot.org] and it seems to be a trend. The fact is that with all the progress made in computing technology, those spinning platters and movable arms have been the bottleneck for some time.

    --Jon
  • This is basically lots of DRAM in a box, right? Not a technical breakthrough.

    A few racks of 1U servers could be configured to have that much DRAM.

    • Difference is..when someone trips on the power cable, this one doesn't lose all of its data. Whereas try doing that in a computer lab during finals week and see if your data in DRAM is gone. I think I remember seeing that Google was using stuff like this for their storage rather than HDD.
  • Memory/RAM (Score:3, Interesting)

    by sean23007 ( 143364 ) on Monday April 08, 2002 @11:15PM (#3307571) Homepage Journal
    So can we expect them to design a new type of system that has non-volatile memory and vast storage in a similar array, divvied up on the fly by the system depending on whether it needs storage or memory at the moment? I've been waiting for the day when memory and hard drive became one, and this seems to take that one step closer to the inevitable.
  • What the heck kind of bus do the expect to drop this wad of data onto? Or are they planning to just map it directly into some memory array? Something like this could change a lot of software - having offline storage faster than main memory is a big deal for many algorithms. The implications are huge! When can I get one and how many lotteries do I need to win?

WARNING TO ALL PERSONNEL: Firings will continue until morale improves.

Working...