Slashdot Mirror


New Device Puts SSD In a DIMM Slot

Vigile points out a new take on SSD from Viking Modular Solutions. The SATADIMM puts an SSD in the form factor of a memory module. "The unit itself actually uses a SandForce SSD controller and draws its power from the DIMM socket directly but still connects to the computer through a SATA connection — nothing fancy like using the memory bus, etc. Performance is actually identical to other SandForce-based SSDs though the benefits for 1U servers and motherboards with dozens of DIMM slots is interesting to say the least. Likely priced outside the realm for average consumers, the SATADIMM will likely stay put in the enterprise market but represents an indicator that companies are realizing SSDs don't need to be in traditional HDD form factors."

169 comments

  1. I suppose the real question here is... by pwnies · · Score: 4, Insightful

    Why? If it's only drawing power from the DIMM slot, what benefit does that serve? Sure, in a 1U rack it *might* save a trivial amount of space. I just dont see a market for it.

    1. Re:I suppose the real question here is... by h4rr4r · · Score: 1

      Add drives to machines that lack enough hard drive slots but have extra dimm slots.

    2. Re:I suppose the real question here is... by Monkeedude1212 · · Score: 5, Insightful

      Sure, in a 1U rack it *might* save a trivial amount of space. I just dont see a market for it.

      If there's anything I've learned from calculus - it's that a whole lot of trivial values can add up to something significant.

    3. Re:I suppose the real question here is... by pushing-robot · · Score: 4, Informative

      I guess it would be a quick way to add storage to a server that has a bunch of unused memory sockets. And the design uses off-the-shelf components which is always nice.

      But there was getting to be a need for a proper SSD package, as sticking them inside HDD housings was both limiting and an inefficient use of space. Viking's solution probably won't take off, though, since Apple/PhotoFast/Toshiba just stole their thunder.

      --
      How can I believe you when you tell me what I don't want to hear?
    4. Re:I suppose the real question here is... by PatPending · · Score: 1

      If there's anything I've learned from calculus - it's that a whole lot of trivial values can add up to something significant.

      That's a good summary.

      --
      What one fool can do, another can. (Ancient Simian Proverb)
    5. Re:I suppose the real question here is... by Anonymous Coward · · Score: 3, Interesting

      Hell yeah, this could save a megaton of space. It seems most of the negative comments are from people who have never seriously used racks

    6. Re:I suppose the real question here is... by tagno25 · · Score: 3, Insightful

      I have, and this would have a hard time fitting in a 1U case. The data cable comes out the top, but many 1U cases have the ram sticks at a 45 degree angle because they would be too tall. It would be OK in a 2U or larger and used as the boot disk.

    7. Re:I suppose the real question here is... by alen · · Score: 1

      i've read where putting the tempdb in MS SQL Server and whatever the Oracle and DB2 equivalents are on SSD is a huge performance boost for queries that rely on it. things like sorts and joins.

      you can easily have multi-terabyte databases on a 1U/2U servers these days and with 16GB DIMM's enough memory in a few slots for them. but if you have idiots running select queries for hundreds of millions of rows at once then this will be a big help. i've seen queries like this run for days

    8. Re:I suppose the real question here is... by silas_moeckel · · Score: 1

      OK lets assume 1ru's come in two basic flavors the fully integrated products from Dell HP IBM and the like these rarely have any standard power connectors let alone internal SATA ports. Then there are the custom built these normally have free standard power connectors and free SATA ports. In the first case there is nothing to plug it into unless you add a sata raid card at which point why not just get the power from the PCI-E slot? Custom servers don't need to draw power from a dimm slot. In either case if your that concerned about performance you would have populated that DIMM slot already to max out ram. This seems like a solution looking for a problem to fix.

      --
      No sir I dont like it.
    9. Re:I suppose the real question here is... by GungaDan · · Score: 1

      If that is true, wouldn't it be better to populate the DIMM slots with RAM and use a ramdisk instead of SSD for this purpose?

      --
      Eloi are stupid, throw morlocks at them!
    10. Re:I suppose the real question here is... by MBCook · · Score: 1

      That was my thought as well. In the article, they seem to have a 90 degree adapter on the SATA cable to plug into the DIMM. My immediate reaction (besides "that's kinda neat") was that RAM is stacked, so if you put 4 of these in a bank of RAM, the 2-4th's SATA cables would hit the cable from the 1st. You'd need cables that connect at 90 degrees in one way and 45 in another.

      If you have empty RAM slots and you want to add one or two, it's not that bad. The idea of using banks of it to put terabytes in a 1U case... seems unlikely.

      --
      Comment forecast: Bits of genius surrounded by a sea of mediocrity.
    11. Re:I suppose the real question here is... by ElectricTurtle · · Score: 3, Interesting

      They should have model variants with connectors staggered relative to the DIMM length. Have one with the connector in the first quarter, another with the connector in the second quarter, etc. So you could have a bank of four with no cable/connector overlap.

      --
      I support the Slashcott and will not be reading or commenting from 2/10/14 to 2/17/14. Beta is steaming pile of dog shit
    12. Re:I suppose the real question here is... by XLazarusX · · Score: 1

      Not just power. It's mounted in a DIMM socket. If your server has lots of DIMM slots and few drive slots this is ideal.

    13. Re:I suppose the real question here is... by phyrexianshaw.ca · · Score: 3, Interesting

      two words for all of you:

      custom cables.

      seriously: sata cables are cheep as hell to build, and doing a fan cable of a custom length to match up to the controller either on board or in the single 16/4x slot would only kind of make sense.

    14. Re:I suppose the real question here is... by sjames · · Score: 1

      Yes, but if it can be that small, just make it about that size but accepting power like a regular drive. Then it can be tucked away anywhere and the cable won't interfere with airflow.

    15. Re:I suppose the real question here is... by phyrexianshaw.ca · · Score: 2, Insightful

      sure, but you know what the price of ram vs the price of flash is?

      16GB dimms run me about $900 each, whereas I can get 64GB X25-E's for $700.
      and tit for tat, the performance won't be THAT bad by comparison.
      at ~$55/GB for Ram, or ~$10/GB for flash, at 1000GB quantities... that's a pretty easy call to make personally. :P

    16. Re:I suppose the real question here is... by h4rr4r · · Score: 3, Interesting

      In a 1U server there no such space. The DIMM design lets you put it in a nice free space and not interrupt airflow too much.

    17. Re:I suppose the real question here is... by phyrexianshaw.ca · · Score: 3, Interesting

      the point is that you can instead of purchasing ram at ~$25/gb you can buy flash at ~$10/gb and still stay dense.

      I'm sure where you are there's room for things: but in much of the world this is not the case. try suggesting 4U storage cases for a customer wanting to host a 20TB database in Egypt. you may only get 4-6U in each building to work with, (with little cooling capacity) and $25K/building in hardware budget.

      There are cases for everything. I can think of a pile of customers of mine that only filled their Vmware hosts with 64GB (of the 512GB max) of ram (leaving twenty eight sockets free in each of the three hosts for something!) that's 33.6TB of space right there! (though personally I'd PREFER to stick RAM in there, that would only be another 1.344TB of ram)

    18. Re:I suppose the real question here is... by sjames · · Score: 1

      I work w/ 1U servers all the time, and there certainly is such space. In the long ones, there's room behind the drive bays, and in a short one, tuck it in to the space between the PCI(e) card (if any) and the MB with a bit of double sticky tape. On older 1Us, put it where the floppy drive used to go.

    19. Re:I suppose the real question here is... by jgagnon · · Score: 1

      And also custom 1U racks filled with powered memory slots for such drives...

      But I seriously think the point to this whole exercise is that with SSD drives we don't have to be tied to any single layout or size... they could be made to go anywhere. They could make them into stackable cubes like Lego's (with sufficient cooling, of course).

      --
      Remember to maintain your supply of /facepalm oil to prevent chafing.
    20. Re:I suppose the real question here is... by Anonymous Coward · · Score: 0

      If that is true, wouldn't it be better to populate the DIMM slots with RAM and use a ramdisk instead of SSD for this purpose?

      With 4 empty DIMM slot, you can install over 1TB of SSD storage. It's not possible to put that much RAM in a DIMM slot yet.

    21. Re:I suppose the real question here is... by phyrexianshaw.ca · · Score: 1

      omfg. Lego storage drives would be awesome.

      who want's to get on that? I'm sure we could find somebody to sell them to.

    22. Re:I suppose the real question here is... by Skal+Tura · · Score: 1

      1U servers lack space for enough HDDs very often, only 3x3½" can ultimately be had. I haven't seen 4x2½" cases neither.

    23. Re:I suppose the real question here is... by amorsen · · Score: 1

      Blade servers. They usually have 2 HDD slots at best. The challenge is that they tend to be low on SAS sockets, so you'd need a very small SAS multiplexer as well.

      --
      Finally! A year of moderation! Ready for 2019?
    24. Re:I suppose the real question here is... by amorsen · · Score: 1

      HP's 1U servers tend to have a couple SATA slots left over, especially if you forego the optical drive (and with PXE or iLO, you don't need it). The actual hard drives tend to run from a SAS RAID controller which often takes up a valuable PCI-E-slot.

      --
      Finally! A year of moderation! Ready for 2019?
    25. Re:I suppose the real question here is... by sjames · · Score: 5, Funny

      You know, there are many decaffeinated coffees on the market that are nearly as good as regular coffee without all the jitters....You should drown yourself in a vat of it.

    26. Re:I suppose the real question here is... by Iamthecheese · · Score: 1

      The MindStorm brick is very hackable and comes with an ARM7 processor. I agree there would be a market for after-market addons for more advanced robotics.

      --
      If video games influenced behavior the Pac Man generation would be eating pills and running away from their problems.
    27. Re:I suppose the real question here is... by soundguy · · Score: 1

      I use Asus RS120-E5 series 1u boxes. Since they're all headless web servers and everything I need is already onboard, there is a considerable amount of unused space above the PCI slots and risers are already provided for power. These boxes only have 4 memory slots though, so in my application, a device built on a PCI card form factor would be a lot more useful.

      --
      Nothing worthwhile ever happens before noon
    28. Re:I suppose the real question here is... by Gilmoure · · Score: 1

      We've secretly replaced the Enterprises dilithium crystals with Foldger's Crystals. Let's see if they notice!

      --
      I drank what? -- Socrates
    29. Re:I suppose the real question here is... by datapharmer · · Score: 1

      3 letters for you: SAN
      If you can't fit your storage into the case you ordered the wrong server or need dedicated storage.
      Use the right tool for the right job. Memory slots are for memory. Servers have extra memory slots because they often need more.
      *gasp* what a concept.

      --
      Get a web developer
    30. Re:I suppose the real question here is... by drsmithy · · Score: 1

      Blade servers. They usually have 2 HDD slots at best. The challenge is that they tend to be low on SAS sockets, so you'd need a very small SAS multiplexer as well.

      If you're trying to put a lot of local storage into a blade server, You're Doing It Wrong.

    31. Re:I suppose the real question here is... by mariushm · · Score: 1

      or just move the connector on a side.

    32. Re:I suppose the real question here is... by Firehed · · Score: 1

      Well, RAM is going to be at least an order of magnitude faster (probably closer to two OOM) if you can find a fast enough bus for it (assuming you mean ramdisks as some sort of probably PCIE-based breakout board, rather than some software-based partition which will get the full speed). Depending on what you're doing, that could be easily worth the extra cost. Anything where you're looking for the performance of a memory-based database table without the actual limitations of using a MEMORY table comes to mind.

      --
      How are sites slashdotted when nobody reads TFAs?
    33. Re:I suppose the real question here is... by Firehed · · Score: 1

      You can fit usually 4 2.5" drives in the space of a 3.5" drive, if you mount them sideways. A special enclosure that routes the drive connectors around the back might make that feasible. Similarly, you could fit twenty* or so of those Macbook Air "blade" SSDs in an appropriate enclosure, giving you 5TB of SSD in the same space of a 2TB spinning disk at current densities. Build in some sort of native RAID5-like controller (hopefully more Drobo-esque with the drive-swapping abilities) and you've really got a killer storage device.

      * a 3.5"HDD is ~100mm wide, the 256GB SSDs are 3.7 mm. Figure a little breathing room plus the physical space for the enclosure (mounting screws,etc) and I figure an average of about 5mm/drive. Maybe that's too dense, but you get the general idea.

      --
      How are sites slashdotted when nobody reads TFAs?
    34. Re:I suppose the real question here is... by afidel · · Score: 1

      I find that hard to fathom, you can fit 8 SFF HDD's in a 1U server which means about 4TB of flash with current densities and about 100k dollars for enterprise class drives.

      --
      There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
    35. Re:I suppose the real question here is... by afidel · · Score: 1

      HP DL 360 G7's can fit 8 2.5" drives.

      --
      There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
    36. Re:I suppose the real question here is... by jameskojiro · · Score: 1

      Who lit the fuse on your tampon?

      --
      Tsukasa: All I really want, is to be left alone...
    37. Re:I suppose the real question here is... by this+great+guy · · Score: 1

      I have worked with dozens of 1U servers from Dell, HP, Supermicro, whitebox, etc, and I have never seen the RAM DIMMs at a 45 degrees angle. Last time DIMMs were at a 45 degrees was 15+ years ago (EDO RAM I think?)

    38. Re:I suppose the real question here is... by WuphonsReach · · Score: 1

      1U servers lack space for enough HDDs very often, only 3x3½" can ultimately be had. I haven't seen 4x2½" cases neither.

      Look here. SDR-S1803-T04 has room for (4) 3.5" disks in a 1U server. Or SDR-S1000-T08 which lets you put (8) 2.5" drives in a 1U space.

      --
      Wolde you bothe eate your cake, and have your cake?
    39. Re:I suppose the real question here is... by inKubus · · Score: 1

      Thank you! This is a stupid idea. First of all, I want *RAM* in my DIMM slots if you're not going to make use of that fast bus.. Secondly, there are already 32 and 64GB microsd flash cards, which are about 1x2cm. So, why not just make a SATA *connector* that is just an SSD, 64-128-256GB. It wouldn't be much larger than the connector itself (think USB keychain flash drive only sata). The only issue is power, but that can be had pretty easily. These things only need 5v I bet, and worst case they need 4 thin wires to carry 1-2W 5 and 12v to them, just make an adapter, super thin daisy chain wire with about 5 or 6 little power connectors and then they plug in on the other side of the SSD. While you're at it, why not just put an SSD "slot" ON THE MOTHERBOARD. Are you serious, they are spending time to make a dimm form factor when they could just as easily design a whole new connector based on existing SATA chipsets and then just have solid state storage right on the motherboard like dimms. And if you're looking for the existing server market, why not just make a self-stick one you can just jam in some corner somewhere? Shit, why not use VELCRO. Anyway, I just got a SATA 90GB Vertex2 (3 1/2" form factor, the actual chips are like 4cm^2, and it's incredible. $2 a GB, but around $2 per MB/s of transfer, which you can't get anywhere else. Don't get me wrong, I want to see more solid state storage, but this DIMM idea is stupid, someone has to stop the children.

      --
      Cool! Amazing Toys.
    40. Re:I suppose the real question here is... by amorsen · · Score: 1

      If you're trying to put a lot of local storage into a blade server, You're Doing It Wrong.

      Why? What is the advantage of using a larger and usually more power consuming form factor?

      For some applications there are no advantages to non-local storage, and the disadvantages can be significant.

      --
      Finally! A year of moderation! Ready for 2019?
    41. Re:I suppose the real question here is... by drsmithy · · Score: 1

      Why? What is the advantage of using a larger and usually more power consuming form factor?

      Because when you spread it across the 16 (or whatever) blades in your chassis, it's not "larger and usually more power consuming".

      For some applications there are no advantages to non-local storage, and the disadvantages can be significant.

      What scenarios are you envisaging where you need high-density computing power and large amounts of local storage ?

    42. Re:I suppose the real question here is... by amorsen · · Score: 1

      What scenarios are you envisaging where you need high-density computing power and large amounts of local storage ?

      I need high density because rack units in shared hosting are awfully expensive and power is even worse.

      If I can move the applications which need (somewhat) large amounts of local storage away from 1U-servers and onto blade servers, I have gained significant density, saved power, and avoided the usual 1U cable chaos.

      --
      Finally! A year of moderation! Ready for 2019?
    43. Re:I suppose the real question here is... by bill_mcgonigle · · Score: 1

      and in a short one, tuck it in to the space between the PCI(e) card (if any) and the MB with a bit of double sticky tape

      Yeah. I tried suggesting something like this to a customer recently. Strapping an SSD to a rail with space under it and grabbing power from the unused PCIe power socket with a field-made cable to solve a system performance problem.

      People freak out about stuff that's not 100% modular. They went with 'accept the performance problem' over buying more U space or hacking in an SSD. With their eyes wide open.

      It's rare that I decide to not take work from a company again, but that one was just too frustrating.

      --
      My God, it's Full of Source!
      OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
    44. Re:I suppose the real question here is... by hitmark · · Score: 1

      Time to teach economists more calculus?

      --
      comment first, facts later. http://chem.tufts.edu/AnswersInScience/RelativityofWrong.htm
    45. Re:I suppose the real question here is... by rsborg · · Score: 1

      Similarly, you could fit twenty* or so of those Macbook Air "blade" SSDs in an appropriate enclosure, giving you 5TB of SSD in the same space of a 2TB spinning disk at current densities.

      This would require a hell of a lot of cooling. Just because it's slim doesn't allow you to ignore heat. I think you might be able to get away with about 2-4x the density of the 2.5" drives, giving you about 2-3TB of space, but the engineering and testing of that would be expensive.

      Ultimately, you will need to bypass the SATA cable as even single SSDs are close to the maximum throughput of SATA2, which means all it would take is roughly 4 SSDs to consistenly saturate a SATA3 bus. That's the reason storage providers have decided to go with PCIe boards ... anywhere you want to put a 3.5" drive, you will likely be able to put a card in instead.

      --
      Make sure everyone's vote counts: Verified Voting
    46. Re:I suppose the real question here is... by drsmithy · · Score: 1

      I need high density because rack units in shared hosting are awfully expensive and power is even worse.

      Most (indeed, all) colos I've ever looked at won't let you put higher density into a rack than can be delivered by 1U servers anyway. They don't have the power and cooling required. So you can use blades, but you won't be allowed to put more than ~40 of them in a single rack.

      If I can move the applications which need (somewhat) large amounts of local storage away from 1U-servers and onto blade servers, I have gained significant density, saved power, and avoided the usual 1U cable chaos.

      What application needs large amounts of high-performance local storage and wouldn't be better serviced by a centralised storage model ?

    47. Re:I suppose the real question here is... by amorsen · · Score: 1

      We find that blade servers make do with less power than 1U servers, per rack. Higher efficiency makes up for the (admittedly modest) density gains. If you avoid spinning disks and go with the Intel Xeon L-series, you can do 16 blade servers in an HP 7000 enclosure on 2kW average.

      And as to centralised storage, you generally lose on latency and cost and rack space. You gain some efficiency and convenience from being able to reassign storage space even without downtime, but it seems silly to place high-performance low-latency SSD's behind a fibre channel or iSCSI fabric. 10Gbps ethernet is also fairly power hungry still, although that will improve.

      Shared storage is great if you actually have a need for it, like if you want to be able to move virtual instances between servers or have multiple servers write to the same file system. Without that need it offers few advantages and quite a few disadvantages.

      --
      Finally! A year of moderation! Ready for 2019?
  2. Just an SSD that uses memory slot power? by rsborg · · Score: 1

    So you don't have to run a molex or other power connector to the SSD, it's easier to put in, I suppose.

    I wonder if there are significant gains to be had by inserting these in place of existing RAM?

    --
    Make sure everyone's vote counts: Verified Voting
    1. Re:Just an SSD that uses memory slot power? by Anonymous Coward · · Score: 0

      I wonder if there are significant gains to be had by inserting these in place of existing RAM?

      Nope, just significant loss

    2. Re:Just an SSD that uses memory slot power? by phyrexianshaw.ca · · Score: 2, Informative

      except storage capacity. they're 512GB (raw, 400GB formatted) per Dimm. making them 32 times denser capacity vs RAM.

      though yes, you're right: compared to the equivalent AMOUNT of RAM, they suck. compared to the same dollar value of ram: that's another story.

  3. Whats the point by headhot · · Score: 1

    If your using a DIMM slot for power, and SATA for data transfer, why not use the power supply for power instead of losing a memory slot?

    1. Re:Whats the point by Monkeedude1212 · · Score: 1

      Are you thinking in a desktop or a server environment? Because I have only ever seen ONE server ever use every single one of its memory slots full of the Max amount of memory for a stick at the time.

      Often times, its trivial to upgrade RAM to get a spare slot.

      It's Not as trivial to have to unplug absolutely everything because you switched out the power supply.

    2. Re:Whats the point by b00m3rang · · Score: 1

      The power still comes from the power supply... where else would it come from? I guess it'd be useful if you have memory slots you're not using, but no extra drive bays.

    3. Re:Whats the point by Rob+the+Bold · · Score: 1

      The power still comes from the power supply... where else would it come from? I guess it'd be useful if you have memory slots you're not using, but no extra drive bays.

      The distinction the GP was making was the power -- yes, from the power supply -- delivered through the pins of the DIMM slot rather than the cable connected directly to the PSU. And I'd have to agree with both of you in asking what the point of this is.

      --
      I am not a crackpot.
    4. Re:Whats the point by Intron · · Score: 1

      If your using a DIMM slot for power, and SATA for data transfer, why not use the power supply for power instead of losing a memory slot?

      Power from a cable has to be regulated to be clean enough to run the flash drive. Motherboard power is already clean and the correct voltage. This saves power regulation and an unneeded drive housing.

      --
      Intron: the portion of DNA which expresses nothing useful.
    5. Re:Whats the point by ArsonSmith · · Score: 1

      Interestingly we don't have any of our ~500 servers that don't have max RAM in them.

      --
      Paying taxes to buy civilization is like paying a hooker to buy love.
    6. Re:Whats the point by Monkeedude1212 · · Score: 1

      They may be using all of their slots - but is every slot full of the Biggest size stick of Ram?

    7. Re:Whats the point by ArsonSmith · · Score: 1

      yes, sort of, there are bigger sticks but they run at slower clock speeds. the Nehalem arcitecture doesn't allow you to run more than 3 busses with i think it was 4 slots each socket (or somthing, just pulling off the top of my head haven't read it in a few months), then in those slots the dim size you put in to max a dual socket system is 96G at 1333 or 192G(ish i forget) at 800mhz and something in between.

      --
      Paying taxes to buy civilization is like paying a hooker to buy love.
  4. Disappointed by wjh31 · · Score: 2, Interesting

    When i saw the headline, i was hoping that this would be a device that allowed an SSD to be connected to a RAM slot and used as RAM, rather than an SSD that takes up a RAM slot.

    Additionally, if they can squeeze a 256GB into a DIMM form factor, why the are even 4GB sticks of RAM still expensive

    1. Re:Disappointed by Rob+the+Bold · · Score: 3, Informative

      Additionally, if they can squeeze a 256GB into a DIMM form factor, why the are even 4GB sticks of RAM still expensive

      Because using flash memory as system RAM would be rather disappointingly slow.

      --
      I am not a crackpot.
    2. Re:Disappointed by Anonymous Coward · · Score: 0

      Additionally, if they can squeeze a 256GB into a DIMM form factor, why the are even 4GB sticks of RAM still expensive

      Is this a joke, or do you honestly not understand the difference between RAM and Flash? Like, they just happen to use the same form factor in this case, one has no bearing on the other. That's like asking why if I can get a PCIe NIC for $20, why do high end graphics cards still cost several hundreds.

    3. Re:Disappointed by TurtleBay · · Score: 1

      A single channel of 1333 Mhz DDR3 RAM can transfer at approximately 11 GB/s. A 240 GB SSD usually has transfer rates of less than 300 MB/s. You pay more for expensive DRAM with an expensive 240 pin dual data rate interface because transfer rate and latency matter that much more in main memory than in storage.

    4. Re:Disappointed by wierd_w · · Score: 3, Interesting

      While the write speed would be painful compared to real DRAM, the read speed would be comparable.

      For large static arrays, and for custom data applications, it could have uses in the form the GP suggests, though it WOULD be a nasty throwback to the days of user ROMs...

      However, I could definitely see the potential in having such a thing mapped directly to system memory, then loading a special block device driver to allocate all that "memory", so that memory IO could be used for data storage. It would eliminate the SATA controller's IO bottleneck, but would impose a slight CPU penalty. For systems with multiple CPUs, that wouldnt be much of a problem. You would need to allocate that memory fast though, to prevent the OS from trying to use it like RAM.

    5. Re:Disappointed by h4rr4r · · Score: 1

      4GB sticks are cheap as hell, kiddo. Even the server stuff as a double kit, meaning 2X4GB is $250.

      How much cheaper can it get?

      And flash as ram would be slow as hell.

    6. Re:Disappointed by DarkOx · · Score: 1

      When i saw the headline, i was hoping that this would be a device that allowed an SSD to be connected to a RAM slot and used as RAM, rather than an SSD that takes up a RAM slot.

       
      Well I don't know why you would want to use slow flash as your primary storage rather than fast DRAM or SRAM.
      Now this is not what I think you had in mind but having primary storage that does not need refresh would permit you to have a machine that could be powered on and off and remain in a consistent state. Well there were a few more things you'd need to do like preserve the content of CPU registers but there are ways to solve those problems. Such a machine also could have only primary storage because large flash banks are so cheap. Now this machine would be very slow but for some applications it would be useful. Actually such machines are often employed in manufacturing and probably other places. A flash based memory module would have been a step toward alowing us geeks how just want to play with something like that a way to build it out of cheap hardware; but your typical gamer would have no interest.

      --
      Repeal the 17th Amendment TODAY! Also Please Read http://www.gnu.org/philosophy/right-to-read.html
    7. Re:Disappointed by surgen · · Score: 1

      Well I don't know why you would want to use slow flash as your primary storage rather than fast DRAM or SRAM.

      Maybe rather than just taking up the slot use it for communication too? Appear as RAM to the computer and then create a RAM drive to mount the SSD? Though it does seems like a rather roundabout way just to avoid using a SATA cable.

    8. Re:Disappointed by phyrexianshaw.ca · · Score: 1

      Sure, burst speed may be comparable: but ram will maintain that burst speed in an almost any number of IOPS. try writing 250KIOPS@1K to a flash drive, watch it slow to a crawl.

      Access times, though much better in flash over the last few years are still an order of magnitude slower. and just imagine writing to ram only to find out that the process must wait while the old blocks are re-allocated due to bad sector remapping. (potentially causing micro or even millisecond access times!)

      sorry, flash has a LONG way to go replacing ram.

    9. Re:Disappointed by Mashiki · · Score: 1

      Depends on if you're using DDR2 or DDR3, DDR2 is dirt cheap. DDR3 is expensive as hell.

      --
      Om, nomnomnom...
    10. Re:Disappointed by amorsen · · Score: 1

      Also, while the problem of flash wearing out has been vastly exaggerated, imagine how quickly a contended lock would wear out the 100.000 write cycles. You could easily do that many in a second, and no wear levelling can cope with that.

      --
      Finally! A year of moderation! Ready for 2019?
    11. Re:Disappointed by m.dillon · · Score: 1

      Typically 2x2G sticks are cheaper than 1x4G stick, particularly when it has to be ECC memory and DDR3. If you are talking about non-ECC memory then you aren't talking seriously. non-ECC memory is just fine for a consumer desktop (though even that is arguable when one is talking about storage in excess of 4GB), but in a server environment ECC is pretty much required. As of about a year ago I've started buying only ECC memory for desktops too.

      Google did a study on memory in 2009, it raised a lot of eyebrows. Let me try googling up some references for people. cnet article. There. That references the pdf too if you want to read the actual paper.

      -Matt

    12. Re:Disappointed by m.dillon · · Score: 1

      It's actually more around 10,000 cycles for consumer grade MLC flash (which is what you find in most SSDs). SLC flash runs around 100,000 cycles. There's been a lot of misinformation on this topic but the easiest way to think about it is to consider the actual write durability that people have been experiencing with SSDs. Take an Intel 40G SSD for example. The vendor-specified write durability is 35TB (I always say 40TB just to make the numbers easy), or 1000x, which assumes a very high write amplification factor of 10x. The ACTUAL durability is closer to 200TB with a 2x write amplification factor when used wisely, which is about what we get with DragonFly's swapcache feature. But it depends heavily on how the SSD is used. Put a normal windblows filesystem on a SSD and you will likely see durabilities more in the range of 40-80TB due to higher write amplification factors and heavier write combining loads on the SSD controller. Put a stupid database which tries to disk-flush every single transaction individually and the write durability will be as low as 40TB. But something like DFly's swapcache which does bulk writes in 64K+ blocks the write durability will be dramatically improved.

      Write amplification is basically due to the fact that a MLC flash chip uses a 128KB write/erase block. Smaller writes either have to be write-combined or otherwise eat a ton more durability due to having to write the whole block than larger writes would. The result is 'write amplification'... more flash is being written to than the actual data being written by the computer. Write combining also requires garbage collection later on... a rewriting which itself creates additional write amplification.

      In anycase, this makes flash totally unsuitable for ram-like activities, but quite suitable for filesystem logs and file data/meta-data caching. Perhaps IBM's M-RAM will be able to fill that niche in a few years. We will just have to see.

      -Matt

    13. Re:Disappointed by Rockoon · · Score: 1

      Write amplification is basically due to the fact that a MLC flash chip uses a 128KB write/erase block. Smaller writes either have to be write-combined or otherwise eat a ton more durability due to having to write the whole block than larger writes would.

      I'm fairly certain that the write block size in every SSD on the market right now is not the same as the erase block size...

      In other words, that 128K block is segmented into 4K blocks (32 of them,) and each 4K block can be written once per erase cycle.

      ..so its not fair to consider small uncombined writes as equivalent to a future erase on a 1:1 basis.. its actually 32:1.. or about 3% of small combined writes will lead to a mandatory erase.

      --
      "His name was James Damore."
    14. Re:Disappointed by Anonymous Coward · · Score: 0

      It would be slow, but still much nicer than a hard drive. No need to serialize objects to and from the hard drive, just leave them sitting in ssd when the power goes down.

    15. Re:Disappointed by drsmithy · · Score: 1

      While the write speed would be painful compared to real DRAM, the read speed would be comparable.

      No it wouldn't. RAM speeds are measured in 10s of GB/sec. SSD speeds are measured in hundreds of MB/sec.

      This is before even getting into the access times, which are similarly disparate.

    16. Re:Disappointed by wierd_w · · Score: 1

      Who said anything about "Replacing" RAM?

      I said it would be handy to have an SSD (as in, permanent storage) mapped directly into the memory bus- Not everything that is mapped into memory is writable, or should be treated like RAM you know. Some memory IO addresses point to nothing, and are used for such various and sundry things as "Virtual memory" space, etc.

      The block driver should look for a signature at the top of the SSD's raw contents, then allocate the whole memory range as a locked, non-paged allocation. That would prevent ANYTHING else from even LOOKING at that memory range, unless your OS is a buggy pile of shit. It would be OWNED by the block device driver. The only process that would have privileges to write to it would be the block device driver's.

      To prevent the OS from frying the SSD with abusive writes (who does that anyway?) the block driver should allocate a resizable write buffer from actual RAM as well, then offload the buffer as CPU load and time permits. If the buffer gets full, the driver should tell the OS to wait.

      Ideally, the offloading process should not occur until after a certain number of miliseconds pass, so that rapidly acessed data can be read from the buffer instead of from the SSD. That way your abusive program's spam of filesystem calls wouldnt kill the SSD.

      Essentially, think "Ramdisk", but with flash memory there, and a read/write buffer allocated by the block driver. NO programs other than the block driver itself should ever try to touch that memory directly; ALL accesses should go through filesystem API calls TO the block driver instead.

      As far as the OS is concerned, the SSD would appear like a disk drive, via the block driver.

      Again, *NOT AS SYSTEM RAM*

    17. Re:Disappointed by petermgreen · · Score: 1

      Typically 2x2G sticks are cheaper than 1x4G stick, particularly when it has to be ECC memory and DDR3.
      Hmm, a quick check of newegg shows the opposite

      Kingston ValueRAM 2GB 240-Pin DDR3 SDRAM DDR3 1066 (PC3 8500) ECC Registered w/parity Server Memory Model KVR1066D3D8R7S/2G $57.98 (including shipping)

      Kingston 4GB 240-Pin DDR3 SDRAM ECC Registered DDR3 1066 (PC3 8500) Server Memory Model KVR1066D3Q8R7S/4G $105.98 (including shipping)

      --
      note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
  5. Reminds of of the old hard cards by bobjr94 · · Score: 3, Interesting

    I remember back before computers had onboard drive controllers and there was no such thing as a standard drive interface, they sold ISA hard drive cards, it was a drive & controller all in one. I dont see to much advantage running a drive on a ram slot, you can just dedicate a drive(s) to you work, swap or temp files. I typically do that when editing video, 1 drive holds the raw videos, one drive is a temp drive and one is what the final video files are outputted to when they are rendered. Much faster then using 1 drive or even a single raid to read/write large amounts of data at the same time.

  6. Huh? by falldeaf · · Score: 0, Redundant

    That's kind of crazy how would you access the drive? Would you need a special driver or something to mount it as a sata drive? This seems neat but I'm confused at how this work and the article didn't seem to explain...

    --
    check out the Mp3 Garbler I built!
    1. Re:Huh? by hjf · · Score: 0, Troll

      FTFS (From the fucking SUMMARY!):

      The unit itself actually uses a SandForce SSD controller and draws its power from the DIMM socket directly but still connects to the computer through a SATA connection — nothing fancy like using the memory bus, etc.

      and the article even have photos of the thing with a SATA cable coming out of it.

      Sheesh!

    2. Re:Huh? by Mitsoid · · Score: 1

      it sounds as if the DIMM slot is simply the "Bay" holding the drive, which also provides the power. Otherwise it will function like an SATA SSD as it connects to the SATA Port for data transfer.

    3. Re:Huh? by AnonGCB · · Score: 0, Troll

      You're an idiot. It's in the summary AND the article, and if you looked even briefly at the actual photo of the device, you'd have seen that it has a sata port on it.

      --
      http://CryoLANparty.com/ A lan I'm staff on!
    4. Re:Huh? by falldeaf · · Score: 0, Offtopic

      Wtf?... I try not to post about bad moderation but how the fuck is that a troll?

      --
      check out the Mp3 Garbler I built!
    5. Re:Huh? by falldeaf · · Score: 1

      Ok now if that's the case this doesn't seem as neat... I guess if you happen to have an empty dimm slot but that seems like a small niche. Why not just sell a tiny ssd drive that you could hook up to the molex power sockets and the sata cable? I'm not a server admin or anything though so maybe I'm missing something.

      --
      check out the Mp3 Garbler I built!
    6. Re:Huh? by PatPending · · Score: 1

      I try not to post about bad moderation but how the fuck is that a troll?

      Do, or do not. There is no 'try.'

      --
      What one fool can do, another can. (Ancient Simian Proverb)
    7. Re:Huh? by Anonymous Coward · · Score: 0

      Yeah, a brain.

    8. Re:Huh? by falldeaf · · Score: 0, Troll

      Yah, kind of a brain fart I guess. In my defense it's kind of a strange niche concept I just didn't get it at first. And since we're name calling you're a small dick, asshole who's mom is a whore and a weasel fucker.

      --
      check out the Mp3 Garbler I built!
    9. Re:Huh? by Anonymous Coward · · Score: 0

      People were being nice and assuming you weren't retarded. ;-)

    10. Re:Huh? by falldeaf · · Score: 0, Offtopic

      That's a really good point, thank you. Let me be more clear. Before, I didn't post about shitty moderation but now I do whatever the hell I want.

      --
      check out the Mp3 Garbler I built!
    11. Re:Huh? by Anonymous Coward · · Score: 0

      Because a server rarely has an extra molex connector, but will often have empty RAM slots.

    12. Re:Huh? by falldeaf · · Score: 1

      Ah, that makes sense, then.

      --
      check out the Mp3 Garbler I built!
    13. Re:Huh? by falldeaf · · Score: 0, Offtopic

      Fuck you , coward.

      --
      check out the Mp3 Garbler I built!
    14. Re:Huh? by phyrexianshaw.ca · · Score: 1

      How many molex connectors are in the average machine?

      there are 32 dimm slots on a modern 4 socket server board. memory comes up to 16GB/dimm densities, and few companies (well, few of the ones I sell to globally) have any need to max the boards. 8 slots gives you 128GB of RAM, leaving 24 slots available.

      at 400GB usable densities per slot with this product, the machine can then host 9.6TB of SSD storage BEFORE connecting to a drive array. being that most cases will only allow 8XSSD's to be mounted, that allows for 8X64GB + 24X400GB = 12.1TB in a 1U box. at that same density, you could cram almost 40TB OF FLASH into the same size as a desktop computer mounted horizontally. (not really, but it gives you the idea. :P)

      the densest thing that I can think of for SSD's, would be a Super Micro SC-417. that's 72x64GB (the largest enterprise SSD's I've ever trusted are the 64GB intel x25-E) yielding a whopping 4.6TB in 4U.

      personally: I see that as pretty awesome.

    15. Re:Huh? by Anonymous Coward · · Score: 0

      You don't get to name-call people for acting normally. Unless you have some kind of organic brain damage, you deserved every bit of scorn and ridicule heaped on you, and now you're here still barking like a loon. Give it up, man.

    16. Re:Huh? by amorsen · · Score: 1

      If you have a molex connector you can easily split it into as many as you want. Flash drives use negligible power.

      --
      Finally! A year of moderation! Ready for 2019?
    17. Re:Huh? by Anonymous Coward · · Score: 0

      I try not to post about bad moderation but how the fuck is that a troll?

      Do, or do not. There is no 'try.'

      Well, by "try" he obviously means that he avoids doing it until the moderation is so blatantly stupid that the urge to post overtakes his personal resistance to post about bad moderation.

      You are either a clueless idiot with no social comprehension, or you are a pedantic idiot looking for something to be a pedant about and feel superior. Neither is very likable. I am going to go out on a limb and assume that you don't have many friends (if any) and most of your co-workers find you difficult to talk to and unapproachable.

      Regards,
      Paul

    18. Re:Huh? by PatPending · · Score: 1
      Paul;

      His OP was (to use your words) "so blatantly stupid" that if he can't figure out why he got a bad mod, well, too bad. He should realize the moderation system is imperfect, as explained here. Specifically:

      I found a comment that was unfairly moderated!

      Lemme know and I'll look at it. Sometimes I might agree and revoke access to a moderator. Usually I disagree and let it go. Its difficult to be the judge on this stuff since it is so subjective.

      His OP was (to use your words) "so blatantly stupid" that my response was too. Tough.

      Your last paragraph is undeserving of comment.

      --
      What one fool can do, another can. (Ancient Simian Proverb)
    19. Re:Huh? by PatPending · · Score: 1
      Paul; I've changed my mind; I will respond to this after all:

      You are either a clueless idiot with no social comprehension, or you are a pedantic idiot looking for something to be a pedant about and feel superior. Neither is very likable. I am going to go out on a limb and assume that you don't have many friends (if any) and most of your co-workers find you difficult to talk to and unapproachable.

      1. Are you assuming that there is a direct correlation between my on-line, anonymous posts on an open forum and my face-to-face, interpersonal relationships? If so, citation needed. (Regardless, it isn't true in my case.)

      2. (Directed to the OP) It's the Internet (in general; Slashdot in particular); if you don't have a sense of humor or are easily offended you shouldn't be on open forums.

      3. I think your diatribe (and--ooooh, you used "pedantic," too!) makes you "feel superior" (your words). Remember, Paul, "When you point your finger, there are three pointing back at you."

      --
      What one fool can do, another can. (Ancient Simian Proverb)
  7. Arima NM46x by Anonymous Coward · · Score: 0

    Something like this would be great with the Arima NM46x, it has 16 DIMM slots.

  8. Speedy servers by $RANDOMLUSER · · Score: 4, Interesting

    Certainly putting things like swap space and database journal files on SSD would speed things up wonderfully, but how about an OS hack where an SSD drive is a sort of L3 cache between core and traditional disk for dirty disk buffers? Also, I'm wondering about the power requirements between SSD and DIMM RAM.

    --
    No folly is more costly than the folly of intolerant idealism. - Winston Churchill
    1. Re:Speedy servers by hjf · · Score: 2

      ZFS L2ARC

    2. Re:Speedy servers by m.dillon · · Score: 4, Interesting

      Not sure I would call it an OS hack but DragonFly has precisely that, called swapcache. Swapcache Manual Page. It isn't so much making standard paging work better (systems rarely have to 'page' these days) but instead its ability to cache clean data and meta-data from the much larger terrabyte+ hard drive that makes the difference. Anyone who has more than a few hundred thousand files to contend with will know what I mean. -Matt

    3. Re:Speedy servers by wagnerrp · · Score: 1

      An SSD usually uses considerably less power even during writing than RAM. Consider that a stick of RAM is going to have to continually refresh each of its 16 or 32 chips, while flash is only going to power up those that it is currently accessing at that time.

    4. Re:Speedy servers by Anonymous Coward · · Score: 0

      If you have RAM slots available why would you stick one of these into it to put your swap space on? when you could just put more physical RAM into the machine which would be so many times faster than the SSD? Not to mention a swap file on a SSD is just asking for premature failure of the SSD

      I seriously don't see the point of this at all. Maybe if the data transfers took place over the memory bus it would be worthwhile, since this should technically be even faster than PCI-E, however i don't even think SSDs are capable of saturating a PCI-E bus yet.

      Seriously what 1U server does not have space for a 2.5" SSD in it? Maybe try moving your existing spindle HDs out of the server and put that storage into a SAN and put your SSDs in place of the spinning disks!

      In its current form where does this really even have a use? In a tiny HTPC box? probably not, ram slots are usually at a premium on those small boards, and there is certainly enough room for multiple 2.5" drives in the average HTPC box. Not to mention on a HTPC box the CPU and RAM are going to make for better performance investments than the HD the OS is booting off. You're going to want a huge disk on a HTPC anyways, unless you also have a NAS or external HD storing all your media

    5. Re:Speedy servers by wuzzeb · · Score: 1

      There is a patch for linux to do this called bcache. See LWN article about it.

    6. Re:Speedy servers by this+great+guy · · Score: 1

      And Linux has bcache.
      And Solaris has L2ARC for ZFS (zpool add XXX cache cXdXtX).

      Both of these technologies allow the OS to uses a fast SSD as a general cache device for data on slower HDDs.

  9. What purpose does this serve? by MobyDisk · · Score: 1

    From the article:

    Final Thoughts: Taking power (and space) from free DIMM slots is certainly a novel idea, and is beneficial to overly cramped installations. I can easily see these being used for embedded and other custom systems where high storage performance is needed without the wasted space.

    So the entire purpose of this hyper-expensive convoluted creation is to save a power cable...? The whole article reads more like an advertisement + some benchmarks. I see no benefit to this thing whatsoever. Unless I am missing something, it sounds more like Viking was trying to make a non-volatile memory chip (that would be kinda cool) and realized it wasn't going to work, so they had the engineers rip out everything novel about it and just use the DIMM slot to save a power cord.

    1. Re:What purpose does this serve? by Anonymous Coward · · Score: 1

      The purpose is to maximize density. Let's say your server has 12 memory slots but you're only using 4 of them. That means it now has 8 free drive bays you can fill up with high-speed SSDs. Have you ever seen a 1U server with 8 drive bays? If you leave off the standard drive bays, it's not hard to make a 1U server with 32 DIMM slots, meaning you can have 32 SDDs in a 1U server and not have to get a power supply with 32 cables or figure out how to route those 32 cables without impeding airflow.

      In theory you could fit 128 DIMM slots on a 1U board, while you can't fit more than 12 HDDs in the same space. Thus, the DIMM form factor gives you a 10x density increase.

      dom

    2. Re:What purpose does this serve? by noidentity · · Score: 1

      I wouldn't be surprised if the purpose were to confuse the buyer. Imagine, you see an SSD that plugs in to a DIMM slot. "Woah, that's got to be faster than normal SSD! Or it's got to be doing something that makes it better than this other one that only connects to a little ribbon cable.

    3. Re:What purpose does this serve? by m.dillon · · Score: 1

      But this makes even LESS sense for large-form-factor mobos. If you want to maximize density you buy a high-density SSD (they come in terrabyte sizes now, after all). You don't buy a hundred custom-fit DIMMs with discrete SATA connectors. It doesn't even make any sense for 1U FF, since 2.5" drive bays trivially fit in that form factor.

      -Matt

  10. kinda dumb by ILuvRamen · · Score: 0

    This definitely falls into the realm of why would anyone do this? It's the same as a regular hard drive but you can't control where it goes as much. Sounds like a heating nightmare. If it's the same performance, then that's the only difference. I suppose a lack of power cables running across the case is a tiny advantage but whatever, this is still pointless.

    --
    Google's Super Secret Search Algorithm: SELECT @search_results FROM internet WHERE @search_results = 'good'
    1. Re:kinda dumb by AnonGCB · · Score: 0, Flamebait

      This isn't aimed at desktops dumbass, this is aimed at servers where iop/m^3 is important

      --
      http://CryoLANparty.com/ A lan I'm staff on!
    2. Re:kinda dumb by DarkXale · · Score: 2, Informative

      You basically completely failed to look at the difference in size, or consider the type of case it is mounted in. Sure its unnecessary in a traditional Tower Case, where there are often 8 hard drive slots available. But in a small server rack where space is premium; this becomes very viable, especially as there are often few drive slots available; but probably there are unused RAM slots.

    3. Re:kinda dumb by afabbro · · Score: 1

      You're an idiot. It's in the summary AND the article, and if you looked even briefly at the actual photo of the device, you'd have seen that it has a sata port on it.

      This isn't aimed at desktops dumbass, this is aimed at servers where iop/m^3 is important

      Is it impossible for you to interact with others without insulting them?

      --
      Advice: on VPS providers
    4. Re:kinda dumb by m.dillon · · Score: 1

      Not sure why you think it would be viable. Small server racks still have very deep footprints and already have plenty of front-loaded 2.5" and/or 3.5" hot-swappable slots. You are advocating that this DIMM thingy would somehow be an improvement? It isn't hot-swappable, it still needs a separate mobo SATA connector (and cable), you actually have to PULL the freaking server out of the rack to change it out. It essentially can't be upgraded. It isn't commodity hardware. AND it is low density compared to what can be stuffed into a 2.5" or 3.5" SSD. Servers are the last place you would ever want to use one of these things.

      The size argument only works if all you wanted was one of these things, and only if space were at a premium. Space is not at a premium in racked servers, not even for 1U. There is always going to be room for 6-8 front-loaded 2.5" standard form factor drives.

      -Matt

    5. Re:kinda dumb by drsmithy · · Score: 1

      This isn't aimed at desktops dumbass, this is aimed at servers where iop/m^3 is important

      In those environments your storage isn't in the local machine anyway, it's in the SAN.

  11. Form factor? by Anonymous Coward · · Score: 0

    What sense does it make to abandon the traditional form factor when it already works with great efficiency?

    I thought that this SSD was actually interfacing with the memory bus, which also would've been a dumb idea given the discrepancies in cost and performance.

    But the headline reminded me of OCZ's DDR booster from years ago. "What next?"

    1. Re:Form factor? by DCstewieG · · Score: 1

      Think Mac Minis or Nano-ITX boards. You could make a damn small box which for many (most?) people is more desirable than expansion room. The case could also be dead simple with the most complicated thing being the holes to attach the board.

    2. Re:Form factor? by m.dillon · · Score: 1

      I think its a bit silly too. For one thing, there is ALREADY a suitable form factor: mini-PCIe. And two, DIMM slots change every year. Anyone buying this dimm-based SSD is basically buying a custom part with no resale value (because its form factor will become obsolete very quickly) and wasting a memory slot that they might actually want to use in the future. Bad news all around.

      -Matt

    3. Re:Form factor? by DarkXale · · Score: 0, Flamebait

      Try saying that again thinking in a context outside of your home desktop.

    4. Re:Form factor? by m.dillon · · Score: 1

      I guess you think a brain-dead one-liner comment like that is meaningful. Try again. I'm sure if you actually spend more than 5 seconds thinking about it you can come up with something better.

      -Matt

    5. Re:Form factor? by h4rr4r · · Score: 1

      Think about the server room is what he was trying to tell you. 1U servers just about always have only two PCIe slots but tons of extra ram slots. Not everything is sold to the consumer market.

    6. Re:Form factor? by m.dillon · · Score: 1

      I think you are misapprehending the density argument. These DIMM slot SSDs are NOT ultra-high-density items. A standard form factor 2.5" or 3.5" SSD is much higher density when you are talking about more than a few gigabytes worth of SSD. They can fit 1TB+ into a 3.5" form factor already. Also, 1U rack mount boxes have no trouble fitting a whole crapload of 2.5" or 3.5" front-loaded slots into the box. It is after all a fairly deep form factor.

      In a server room these DIMM SSDs are the last thing you would ever want to use. Not hot swappable, low density, custom-fit hardware. I wouldn't touch these things with a ten-foot pole.

      -Matt

  12. PCIe RAID card? by BC_R3 · · Score: 1

    Correct me if I'm wrong but couldn't all of this be done with a PCIe RAID card?

    1. Re:PCIe RAID card? by DarkXale · · Score: 1

      Considering smaller form factor machines generally only have one or two PCI-E slots, and these aren't infrequently used for Network connectivity (or GPUs) - chances are high you simply don't have any PCI-E slots left.

  13. Because RAM isn't Flash by Sycraft-fu · · Score: 3, Informative

    The price of flash has nothing to do with the price of RAM. They are completely different constructions, and for different tasks. Flash is faster than magnetic storage but still dog slow compared to RAM. For flash you talk access time in 2-3 digits of microseconds. For RAM you talk access time in single digit nanoseconds. For flash transfer rates are in the 100s of MB/sec with anything over 200 being rather exceptional. For ram transfer rates are in the 10+GB/sec.

    Same sort of transition again when talking DRAM (what you put in your system) to SRAM (what processor cache is made out of). Again the price goes up massively so instead of 8GB, you are talking maybe 12MB. However again the speed goes way up and access time way down.

    1. Re:Because RAM isn't Flash by Anonymous Coward · · Score: 0

      For flash you talk access time in 2-3 digits of microseconds.

      Are you sure about that? A $1 microcontroller can fetch single cycle instructions from flash at 20MHz (=50ns). I don't think you will have any trouble finding one that works at 30ns or even lower.
      I have always assumed that the reasons SSDs are so slow is because of the limits of the SATA-bus and immature controllers. Flash memory by itself does not seem that slow to me.

    2. Re:Because RAM isn't Flash by hamster_nz · · Score: 1

      Mostly correct, but flash isn't THAT slow. My nice little Digilent FPGA board has 70ns flash on it - I can read read a location in under a 10th of a microsecond.

      The latency on SSDs isn't due to the flash itself, but due mostly to the time taken for the data to be squirted up the SATA interface. That is why flash on PCIe cards making an appearance in high-end systems.

    3. Re:Because RAM isn't Flash by m.dillon · · Score: 1

      This is kinds like comparing apples to oranges. Think of the SSD as another intermediate fully random-accessible cache layer that is slower than ram but much faster than a hard drive. Consider the cost of placing, say, 40G of ram in a server. That's a lot of DIMM slots, a more expensive mobo, lots of expensive high density dram, verses the cost of a 40G SSD ($115 from Intel). So even though the SSD is more expensive per gigabyte than a normal HD, it is considerably less expensive per gigabyte relative to dram and far denser. Perfect for another cache layer.

      Access times through a standard SATA interface are not all that bad either. Certainly NOT 3-digit microseconds. Random read access time is something like 38uS with SATA-II @ 26K IOPS with the cheapest Intel SSD. That isn't bad at all for access to (~40G-160G) worth of SSD per SATA channel.

      Much faster custom PCIe card based SSDs exist but they are really only useful for dealing with database transactions. System performance (sans database transaction tests) is going to be fairly agnostic to any access time less than around 100uS. The more important feature is the full random access capability of the SSD verses a normal HD. HD performance goes to hell when it has to be accessed randomly... as in 6mS hell. 38uS vs 6ms... hrmmm.

      -Matt

    4. Re:Because RAM isn't Flash by Anonymous Coward · · Score: 0

      Sounds like you don't know the difference between NOR and NAND flash to me.

  14. because there's no "read the summary" moderation? by Chirs · · Score: 0, Offtopic

    Duh....

  15. saves space primarily by Chirs · · Score: 2, Informative

    It's aimed at 1U servers that have no free drive bays or PCI slots.

  16. Mini Options! by Falc0n · · Score: 4, Interesting

    Actually I find this potentially quite cool. Not as much for the power source, but the size. Since most mATX boards don't come with mini PCIe slots, if you want to use an SSD drive you need a 2.5" drive or a PCIe card with a mini-slot on it. Both are much larger than a DIMM option.

    And with 50gb, this would be very useful in a media box streaming from a server. Now only if the price could come down.

    1. Re:Mini Options! by m.dillon · · Score: 1

      But not more useful than actually putting ram in that slot, since most small form-factor motherboards are also going to have a minimal number of DIMM slots anyway. One rarely sees more than 4 slots and I don't know about everyone else but I always populate all my DIMM slots so I don't have to purchase ultra-high density sticks (which cost a premium).

      There is a very good reason why Apple had to use this sort of thing... a custom-fit SSD in a custom, basically non-upgradeable item (people who buy Apple stuff tend to replace the machines en-toto and not do piecemeal upgrades). Since custom-fit essentially makes the stuff non-upgradeable it just isn't suitable for any other environment. One might not care so much when one purchases the item initially but I guarentee that people will start to care very much when, a year or two down the line, they find the part no longer available and their hardware now worthless because it can't be upgraded.

      So for Apple, it may be a good thing, but sold as a general consumer item it is going to become a has-been very quickly. Except for the ultra-thin market a 2.5" SSD (e.g. most of the Intel SSDs) is plenty small enough. If there were actually demand for this sort of thing we will also see case makers react very quickly and start pushing 2.5" slot-only cases (especially now that the CD/DVD form factor is starting to disappear in favor of USB).

      -Matt

    2. Re:Mini Options! by cyclomedia · · Score: 1

      Read GP again, if you're building a home cinema box and are using a small form factor mobo, and that mobo has (for example) two dimm slots, stick 1GB ram in one slot and a 50GB SSD drive in the other slot and you're done, no extra space/cooling and bulky aluminium mounting for a traditional HDD needed. Very cool (pun intended)

      --
      If you don't risk failure you don't risk success.
  17. Why so many ignorant replies? by PatPending · · Score: 2, Insightful

    I can't recall a /. story that has this many ignorant replies.

    Aside from the usual lack of RTFS and not reading TFA, I wonder if it's due to ignorance of hardware?

    --
    What one fool can do, another can. (Ancient Simian Proverb)
    1. Re:Why so many ignorant replies? by falldeaf · · Score: 1

      I'll just speak for myself but to me it seems like a strange concept to use a memory slot for power... apparently memory slots are commonly abundant in servers but not molex cables? I guess niche concept and 20 acronyms leads to mistake making territory in my case.. whatevs....

      --
      check out the Mp3 Garbler I built!
    2. Re:Why so many ignorant replies? by Anonymous Coward · · Score: 0

      Slashdot is a great and glorious cluster fuck of Internet wisdom. About 99% of the comments are from know-nothings that think they know or are trying to bullshit you. Most of the articles are rehashes of other technolgy sites (such as Ars) or stale general news stories. I can drop Slashdot and never miss any news. Slashdot, like Fucked Company, is long past its prime.

      The only reason I come here is to shake my head in disbelief at the level of stupidty and point out that these people have made the IT business the shithole it is today.

    3. Re:Why so many ignorant replies? by Jeng · · Score: 1

      And don't forget that often times it is people such as myself modding the conversation. I try my best to moderate well, but on a subject like this I am outside of my knowledge base. I think that is a reason that so few comments on this story are modded much at all, besides the quality of some of the posts are just saying wtf without bothering to reading the article.

      Slashdot does seem a bit desperate for moderation, I'm getting 30 mod points on some weeks, 15 at a time.

      --
      Don't know something? Look it up. Still don't know? Then ask.
    4. Re:Why so many ignorant replies? by MightyMartian · · Score: 0, Flamebait

      Translation: I'm an incredible prick. My family hates me. My friends hate me. Hell, even the goddamned cockroaches feasting on microwaved mac-and-cheese I've left rotting in the sink hate me. I think I'll go to Slashdot and show everybody how outrageously awful I really am.

      --
      The world's burning. Moped Jesus spotted on I50. Details at 11.
  18. Distinguishing users from admins by Anonymous Coward · · Score: 0

    These comments give everyone the chance to figure out whom works with computer hardware and whom just owns a desktop.

  19. Useless with virtualization? by ferrocene · · Score: 4, Insightful

    This device seems backwards with today's trends. With virtualizaion gaining ground fast, the ideal setup is to have as much RAM as possible with a SAN back end for storage - iSCSI, FC, whatever. Most local disks on servers today are RAID1 mirrors for the small hypervisor.

    So, yes, this device wastes a valuable DIMM slot to give you a less-valuable SATA drive?

    I can't think of any scenario where this would be useful unless you're talking about handheld devices - a MacBook Air or tablet of some sort.

    --
    Most folk'll never lose a toe, and then again some folk'll...
    1. Re:Useless with virtualization? by h4rr4r · · Score: 2, Insightful

      DB servers in a leased rack. Doing DB IO over FC or iSCSI adds latency that local disks are not going to have. This gets you fast local storage without having to pay more each month for leased rack space.

      Virtualizing high performance DBs is a stupid move.

    2. Re:Useless with virtualization? by Anonymous Coward · · Score: 0

      Are you saying that a FC SAN will give you fewer IOps than this DIMM SSD?

    3. Re:Useless with virtualization? by chgros · · Score: 1

      Are you saying that a FC SAN will give you fewer IOps than this DIMM SSD?
      He's talking about latency, not throughput.

  20. Rack? by zmooc · · Score: 1

    Rack? Who cares about racks. It's not like there's not enough room in 1U servers. What this is awesome for, though, is for small form factor PCs. With video on the mobo or cpu the only thing left that stuck out, was the harddrive or ssd. Not anymore. Awesome! :-) Now I can go get myself a proper 17x17x5cm quad core PC:-)

    --
    0x or or snor perron?!
  21. So... wait. by sammy+baby · · Score: 1

    I can see this in an environment when you need to stick a lot of 1U rack systems all over the place, and can't spread out over a larger footprint in any one location. But when else am I going to use this? Didn't we decide a long time ago that large amounts of internal storage wasn't really a good way to handle increasing storage needs?

    I'd much rather see a big ol' SAN full of SDDs than put together something like this, unless someone else is seeing an advantage that I don't.

    1. Re:So... wait. by Anonymous Coward · · Score: 0

      If you need a lot of storage and a lot of processing power, and if you can use distributed storage like hadoop or luster, then SSD in DIMM slots could increase your storage capacity for each node by a TB or more.

  22. Wow, someone put a piece of hardware ... by Kaz+Kylheku · · Score: 1

    ... into a different electronic orifice of the motherboard than what is customary?

    This is exciting news, indeed!

    I will join this game-changing revolution by using file descriptor 3 for standard output!

  23. We need a new standard. by Above · · Score: 1

    We need a new standard form factor or two. Clearly making an SSD in the size of a pattered hard drive makes no sense, but this product makes no sense either. It's just a way to steal power from another sort of slot. In addition to the form factor, I'm not sure SATA even makes sense anymore, so it may be time for a higher level rethink.

    I'm not sure the best way to go, but there are some semi-obvious starting points. What about MiniPCI for SSD's? One or two on the motherboard could work well. Maybe a modified SATA design with the cable suppling power for SSD's? They could look like mini-USB sticks and plug directly into the connectors. Maybe we need an entirely new bus interface just for SSD's, and then put them in a SIMM/DIMM type package; that may be the best option for performance and size.

    1. Re:We need a new standard. by Daniel_Staal · · Score: 1

      The new formfactor introduced in the Macbook Air sounds interesting, and already announced to be available from a couple of different manufacturers. It's basically the same size as a DIMM, but with the pins at the end instead of along one edge.

      --
      'Sensible' is a curse word.
  24. Just not practical due to cabling by Anonymous Coward · · Score: 0

    This idea simply does not scale due to SATA cabling limitations, especially in existing systems that locate their DIMM slots for proximity to the CPU, not the onboard SATA controllers. Even a relatively low density of SSD DIMMS would result in a mess of cabling in a typical case let along that crazy 32DIMM/row picture they provide.

    Notice they never note the height. Since the SATA connector is in the same location on all the DIMMS and DIMM slots are traditionally lined up any use of standard SATA connectors requires them to be stacked. That means in that crazy picture with 128 slots you would have 32 stacked SATA cables and connectors. Even at 2mm thick that is 64mm, or 2.5" of sata cable all stacked. That is probably best case and that does not even take into account the mess at the controller having to plug 32 SATA cables into 32 SATA ports. There is a reason that higher density disk systems use backplanes that provide power and data connectivity.

    The only interesting thing about this product is that you can put SSD chips onto almost anything. Far more interesting would be an industry standard form factor that allows for such densities but where the controller wiring is in the slot as well. Then you can integrate a proper solution like SAS to SAS/SATA SSD Chip Slot backplane so your cabling is not a complete mess.

    The new form factor for the new thin apple laptop is much more promising.

  25. Re:because there's no "read the summary" moderatio by falldeaf · · Score: 0, Offtopic

    Ah, well I did read the summary but I didn't get it I guess. Still, over rated maybe? I just made a mistake, I wasn't trying to be inflammatory or troll anyone.

    --
    check out the Mp3 Garbler I built!
  26. Reasonable packaging by Animats · · Score: 1

    It's not a very exciting use of non-volatile memory. It makes sense, though, to package non-volatile devices for vertical slots like DRAM, and have motherboards that have slots for them. But not DIMM slots - something that actually carries the drive data. The thing announced in the article still needs a drive cable; all it gets from the DIMM slot is power. This looks like an interim product until server motherboards go to that form factor and eliminate drive bays. The near future for server farms probably looks like that - onboard non-volatile storage in some kind of vertical slot, with rotating disks elsewhere in a storage array.

    One of the challenges in computing is to figure out what to do with non-volatile memory besides pretending it is a "disk". Today, CPUs and operating systems understand two kinds of storage - "RAM" and "disk". Design has been locked into that model for decades. Nobody really knows what to do with something that has a 35us access time and no variable latency. Going through the operating system's file system drivers runs up the latency, but making big devices accessible as memory makes them too vulnerable. Some kind of intermediate form of access is needed. A tuple store? A database implemented in an FPGA? Something like that might make sense.

    1. Re:Reasonable packaging by emj · · Score: 1

      actually quite a few embedded platforms use flash as swap.

  27. Gigabytes per gram by wazzap123 · · Score: 1

    There's an interesting analysis over at the daily circuit that covers this and the new Toshiba form factor, comparing in terms of Gigabytes per gram with 640GB 2.5" and 3TB 3.5" drives.

  28. New? by Lumpy · · Score: 3, Funny

    I've had such things in the embedded world for over a decade.

    What's next? NEW! small cards serve as memory devices!

    --
    Do not look at laser with remaining good eye.
    1. Re:New? by Overzeetop · · Score: 1

      It'll never happen. Floppy and CDs work just fine in digital cameras. Why would you even want something so small you might lose?

      --
      Is it just my observation, or are there way too many stupid people in the world?
  29. Who has empty DIMM slots? by Anonymous Coward · · Score: 0

    We use blades even (smaller than 1U) and we run out of DIMM slots before we run out of CPU or I/O.

    What kind of application needs lots of servers each underpopulated with RAM and also populated with their own SSD?

  30. You can get RAID controllers that do that by Sycraft-fu · · Score: 1

    Adaptec's like of MaxIQ controllers are the cheapest I know of, Intel also has it on their high end rebadged LSI controllers though you have to pay extra to add the feature. The controllers use an SSD as an additional layer of cache (they also have a RAM cache) for the array to speed things up. Works quite well apparently, if a bit costly.

  31. Sacrifice memory for storage space? by gravis777 · · Score: 1

    Let me get this straight, you want us to sacrifice valuable RAM slots, and more so, valuable RAM, to run an SSD device? What would make more sense would be to have a completely seperate 1U unit hooked up to the unit with nothing but SSD devices (or hard drives). Wait, don't they already have those?

    Likely priced outside the realm for average consumers,

    I also doubt the average consumer will want these. With most consumer motherboards only supporting two or four slots of RAM, I REALLY don't see sacrificing ram slots for SSD. Especially when they top out at, what, 128gig? I just bought a 2TB harddrive for $94. I mean, I guess I could put a single 4gig memory chip in my machine and three of these, but this gives me, what, like 378 gig of space?

  32. SSD as memory by Anonymous Coward · · Score: 0

    Dang, I hoped someone had finally figured out that SSDs would make great (if slightly slow) memory. Why does an SSD need to be connected to SATA? Why not directly to the memory bus?

  33. PrestoServe NVRAM, circa 1996 by bolthole · · Score: 1

    Everything old, is new again.
    Sun did this in the sparc 10 &20 line, by enabling an optional NVRAMM SIMM in the primary memory slot.
    A whopping 4 megabytes of RAM max, I think :) so it was used for caching the "metadata" of things like NFS, rather than direct storage.
    But putting something directly in memory, and accessing it through the memory bus 'normally' like a basic RAMdisk) sounds a whole lot more efficient than just sucking power from the slot, but looping back around through the SATA bus, so you can then get the data... back into the memory area :-/

    http://docs.sun.com/app/docs/doc/802-5670/6i9dnsnct?l=en&a=view

  34. Indeed by cyclomedia · · Score: 1

    No spindle noise too and it's not just the HDD that takes up a bunch of space, the bay itself can usually be quite chunky, and then there's the heat that they generate

    --
    If you don't risk failure you don't risk success.
  35. Useless. by RichiH · · Score: 1

    You still need to use SATA cables. Notice how the picture in the article shows only one of those things in a slot? Guess why.

    Also, for 1U, you usually slant the RAM at 45 degrees. Not possible with straigth nor with angled SATA connectors.

    Also, the idea is ancient.

    But all of this is irrelevant, anyway. People are already working to get the form factor Apple used in the new MacBook Air into other hardware. I am not sure if it's the perfect form factor, but it seems to work reasonably well, it's not a horribly broken crutch like the above and it exists. In 3-12 months, most 1U servers will have one or two of those slots on-board. Problem solved.

  36. citation please? by Anonymous Coward · · Score: 0

    Where have you seen this before?

    I've seen IDE/PATA flash drives that fit inside of the 40-pin connector and can be powered off the IDE pins or externally through a molex connector. And internal USB flash drives that fit on the 5x2 pin mobo connector (which generally fall out during shipping, bad idea). But for SATA the equivalent didn't exist, because SATA can't come up with enough power on its own to drive anything useful.
    It is possible to make a super low profile PCIe flash card that fits in a 1U mobo without any bracket, which would be equivalent to this SATADIMM thing, but better. but all of the PCI/PCIe flash drives on the market are currently full height.

    1. Re:citation please? by Lumpy · · Score: 1

      Here's your citation...

      Posting on slashdot without a sense of humor or guits to post under your account.

      it will cost you 40 karma points for your fine, Judge taco will take bribes to make it go away though.

      --
      Do not look at laser with remaining good eye.