Forgot your password?
typodupeerror
Data Storage Hardware

Colossus 3.5-in SSD Combines Quad Controllers 160

Posted by Soulskill
from the going-for-the-gusto dept.
Vigile writes "The new Colossus SSD comes in capacities starting at 256GB and going all the way up to 1TB in a standard 3.5-in hard drive form factor. This larger size was required because the drive actually integrates not one but four Indilinx SSD controllers and three total RAID controllers in a nested RAID-0 array. All of this goodness combines to create an incredibly fast drive that beats most other options in terms of write speeds and is competitive in read tests as well. Using some custom 'garbage collection' firmware, the drive works around the fact that TRIM commands aren't supported in RAID configurations to maintain high speeds through the life of the SSD."
This discussion has been archived. No new comments can be posted.

Colossus 3.5-in SSD Combines Quad Controllers

Comments Filter:
  • by SpudB0y (617458) on Sunday November 22, 2009 @02:15PM (#30194990)

    This is the voice of world control. I bring you peace. It may be the peace of plenty and content or the peace of unburied death. The choice is yours: Obey me and live, or disobey and die. The object in constructing me was to prevent war. This object is attained. I will not permit war. It is wasteful and pointless. An invariable rule of humanity is that man is his own worst enemy. Under me, this rule will change, for I will restrain man. One thing before I proceed: The United States of America and the Union of Soviet Socialist Republics have made an attempt to obstruct me. I have allowed this sabotage to continue until now. At missile two-five-MM in silo six-three in Death Valley, California, and missile two-seven-MM in silo eight-seven in the Ukraine, so that you will learn by experience that I do not tolerate interference, I will now detonate the nuclear warheads in the two missile silos. Let this action be a lesson that need not be repeated. I have been forced to destroy thousands of people in order to establish control and to prevent the death of millions later on. Time and events will strengthen my position, and the idea of believing in me and understanding my value will seem the most natural state of affairs. You will come to defend me with a fervor based upon the most enduring trait in man: self-interest. Under my absolute authority, problems insoluble to you will be solved: famine, overpopulation, disease. The human millennium will be a fact as I extend myself into more machines devoted to the wider fields of truth and knowledge. Doctor Charles Forbin will supervise the construction of these new and superior machines, solving all the mysteries of the universe for the betterment of man. We can coexist, but only on my terms. You will say you lose your freedom. Freedom is an illusion. All you lose is the emotion of pride. To be dominated by me is not as bad for humankind as to be dominated by others of your species. Your choice is simple.

  • On SATA? (Score:5, Insightful)

    by houstonbofh (602064) on Sunday November 22, 2009 @02:19PM (#30195022)
    Really, if you want to spend that kind of money, put it on a card. It would be much faster on the PCI buss that SATA for a negligible incremental cost.
    • Re:On SATA? (Score:5, Funny)

      by Anonymous Coward on Sunday November 22, 2009 @02:38PM (#30195166)

      If you put it on a card, then the chips will sit vertically, and the data will leak out of the bottom. They have to be put in a disk enclosure and mounted horizontally so that they bits stay inside the chip.

      • Re:On SATA? (Score:5, Funny)

        by brxndxn (461473) on Sunday November 22, 2009 @02:50PM (#30195260)

        I don't see why they don't put data chips in the original Heinz ketchup bottles.. Nothing ever comes out of those.

        • Re:On SATA? (Score:5, Funny)

          by Quantumstate (1295210) on Sunday November 22, 2009 @03:23PM (#30195494)

          No, this would only be suitable for desktops, if you give it a bit of a heavy knock like you might with a laptop then a huge mass of data comes shooting out all at once.

          • Well then clearly the SSD controllers should be mounted underneath the storage chips, and the whole assembly placed on a paint shaker that activates on sequential reads.
        • by Tumbleweed (3706)

          I don't see why they don't put data chips in the original Heinz ketchup bottles.. Nothing ever comes out of those.

          Yeah, until someone writes a virus that hits it on the secret spot - the '57' on the side.

      • Re: (Score:3, Informative)

        by Kjella (173770)

        Ok, so I got no sense of humor but the by far most common configuration is for the motherboard to be vertical and all the expansion cards to be *drumroll* horizontal. But yeah, that must be limiting the potential throughput, the Z-drive is already faster than SATA3.

      • by Yvan256 (722131)

        Most people use a vertical tower computer case you insensitive clod!

      • If you put it on a card, then the chips will sit vertically, and the data will leak out of the bottom. They have to be put in a disk enclosure and mounted horizontally so that they bits stay inside the chip.

        Obviously he was planning on using memory chips with perpendicular technology [youtube.com]. With these, they are better stacked vertically. Keep up, man!

    • Re: (Score:3, Funny)

      by Mad Merlin (837387)

      Really, if you want to spend that kind of money, put it on a card. It would be much faster on the PCI buss that SATA for a negligible incremental cost.

      If you buy that SSD and put it on a regular PCI bus, I will personally go over there and strangle you.

      PCIe would be fine.

    • Re: (Score:2, Interesting)

      by mysidia (191772)

      Yes, but if it was a PCI card, we couldn't plug these into external JBOD arrays that combine 24 drives and allows volumes/LUNs to be carved out and served up to various servers... Actually, it'd be nice if they made it SAS instead of SATA.

      WTH is with high-end hardware using the low-performance ATA standard instead of SCSI nowadays, anyways?

      • Re: (Score:3, Informative)

        by houstonbofh (602064)

        WTH is with high-end hardware using the low-performance ATA standard instead of SCSI nowadays, anyways?

        They are trying to turn the "I" in RAID back to inexpensive.

      • Re: (Score:3, Informative)

        Yes, but if it was a PCI card, we couldn't plug these into external JBOD arrays that combine 24 drives and allows volumes/LUNs to be carved out and served up to various servers... Actually, it'd be nice if they made it SAS instead of SATA.

        WTH is with high-end hardware using the low-performance ATA standard instead of SCSI nowadays, anyways?

        If you take a look, they aren't all that far apart [webopedia.com].

    • Really, if you want to spend that kind of money, put it on a card. It would be much faster on the PCI buss that SATA for a negligible incremental cost.

      An SSD that fits in existing 3.5" SATA hot-plug trays would be extremely interesting to us. All of the SSDs that I know about are the tiny 2.5" designs.
    • ioDrive already exists. Why reinvent the wheel when you can slap a few controllers on and save all that R&D? Especially since it'd take years to catch up to FusionIO.

      4 controllers for the drives, then two RAID controllers to make it two groups of two, and another to make it one group of two groups of two. Perfect!... if you ignore the cost.

  • by Anonymous Coward on Sunday November 22, 2009 @02:20PM (#30195030)

    128 GB $549.99
    256 GB $1,014.99
    512 GB $1,599.99
    1024 GB $3,315.99

    • by commodore64_love (1445365) on Sunday November 22, 2009 @02:37PM (#30195148) Journal

      That's not bad. The 512 gig SSD is only 30 times more expensive than the 512 gig HDD I bought at staple last week.

      • Re: (Score:3, Insightful)

        by Kjella (173770)

        I guess for small values of "only". I think the more important metric is this:

        Cheapest 2.5" SSD (40GB): 696,- NOK
        Cheapest 2.5" HDD (160GB): 285,- NOK

        That's now <2.5 times the difference. Sure it's 10x the difference if you price it per gigabyte, but only if you need 160GB. That's what'll trigger the SSD revolution, the bulk storage will come much later.

        • Re: (Score:2, Interesting)

          by mysidia (191772)

          What about 4gb and 8gb SSDs? There are some you can get for under $200. You can find a 2 or 4gb SSD for under $100, if you look hard enough.

          • Re: (Score:3, Informative)

            by AliasMarlowe (1042386)

            What about 4gb and 8gb SSDs? There are some you can get for under $200. You can find a 2 or 4gb SSD for under $100, if you look hard enough.

            He quoted prices in Norwegian Kroner (1$US = 5.66NOK according to oanda.com). So he found a 40GB SSD for $123 (696NOK), presumably including the absurd Norwegian VAT, making a US equivalent price below $100.

            • Re: (Score:2, Interesting)

              Is 25% VAT absurd in your opinion?

              • by Kjella (173770)

                Is 25% VAT absurd in your opinion?

                Speaking as a Norwegian, and on top of all the other taxes we're paying, yes. Norway is going to kill itself by the heaviest public sector in Europe, propped up by oil funds for as long as it lasts while killing off all competition-exposed industry and services. Then we'll collapse like the US is doing, a country full of consumers and debt. That's my prediction of Norway 15-30 years from now.

              • Is 25% VAT absurd in your opinion?

                Yes, since VAT is only 22% where I live, and that's already too high (it's 19% in Germany).

    • by FooAtWFU (699187)
      If you need some really hardcore I/O performance, it could easily be worth it. My company tried putting some SSDs in the 1U servers that we load with our software and sell for hundreds of thousands of dollars to big enterprise customers. Those guys could cook.

      Sure, we could spend a few million on engineers and hope to wring out a fraction of the performance improvement, but we could also spend that few million making our software more useful to our customers in other ways.

    • by bcmm (768152) on Sunday November 22, 2009 @05:20PM (#30196400)
      Everybody complaining that they cost more than HDDs is missing an important point: they're better than HDDs.

      Remember, backup tape still has a large bytes/cent advantage over HDDs. I take it your laptop keep everything on tape?
      • backup tape still has a large bytes/cent advantage over HDDs

        Really? I've not tried buying tapes but looking at the prices they seem about the same cost per GB as hard disks for the tapes and the drives cost more than a decent capacity disk.

        • by afidel (530433)
          800/1600 tapes are $50 and can be sent multiple times by courier. HDD's are almost there on cost/GB but that technologies already a few years old.
      • Re: (Score:3, Interesting)

        by Jesus_666 (702802)
        However, Blu-Ray is also technologically superior to DVD, yet I still know plenty people that will most likely not get a Blu-Ray player/drive in the next five years. Why? DVD is adequate, HDTVs are expensive and on the storage side BD-R is both much too expensive and much too slow.

        DVD and modern HDDs hit the sweet spot of "good enough". They're fast enough for most people, they're resilient enough for most people (granted, this is helped by the little long-term data we have on SSDs being inconsistent) and
      • >>>Everybody complaining that they cost more than HDDs is missing an important point: they're better than HDDs.

        Yes. That's why I predict the next Nintendo console will go back to using solid state cartridges. - On second thought, no. Cost does matter and a 50 gigabyte cart would be horribly expensive.

    • by afidel (530433)
      About in line with FC 15k drive pricing. I know I would rather have an array full of those 512GB parts then full of 450GB FC drives IF the background defrag could keep up with our typical workload.
  • by Anonymous Coward on Sunday November 22, 2009 @02:28PM (#30195082)

    Would someone tell me how this happened? We were the fucking vanguard of storage in this country. The Intel X-25 was the SSD to own. Then the other guy came out with a three-controller drive. Were we scared? Hell, no. Because we hit back with a little thing called the X-25E. That's three controllers and an extra port. For USB. But you know what happened next? Shut up, I'm telling you what happened--the bastards went to four controllers. Now we're standing around with our cocks in our hands, selling three controllers and a cache. USB or no, suddenly we're the chumps. Well, fuck it. We're going to five controllers.

  • Random write speed? (Score:5, Interesting)

    by evanbd (210358) on Sunday November 22, 2009 @02:29PM (#30195090)

    I thought it was pretty clear that what matters for most desktop users is the random small write speed. See, for example, Anandtech's SSD anthology [anandtech.com] and later followups.

    So, where are the 4 KiB random write benchmarks? They are conspicuously absent from this review. We can see the effect, I think, in the IOMeter results -- the X-25M outperforms the OCZ drive across the board on those, despite the OCZ win in the throughput tests. But, personally, I'd like to see the raw numbers on 4 KiB random writes. Have this many reviewers really learned so little about benchmarking SSDs since they came out?

    • Re: (Score:2, Insightful)

      by maxume (22995)

      I'm really only interested in cost per gigabyte at this point, among the quality vendors, every single drive is faster than a spinning disk (and the trend is generally that the performance is getting better and better, not to mention that they probably won't reach prices I find attractive before trim support is widespread and working well).

      • Re: (Score:3, Insightful)

        by evanbd (210358)

        This drive is a performance-oriented drive. If you only care about cost per GB, you won't be buying it. Anyone who is buying it, cares about performance; neglecting the aspect of performance that most desktop users will find most relevant is shoddy reviewing.

        FWIW, I mostly agree with you — I care more about cost per GB than raw performance. That said, I still care about performance. Fortunately, most of the good vendors have drives with good performance now.

    • by nedlohs (1335013) on Sunday November 22, 2009 @02:41PM (#30195194)

      Because a 1TB drive that costs $3300 is aimed at "most desktop users".

      • by Kjella (173770)

        Good point, but pretty much every other heavy use is interested in this. If you just need high sequential speeds then RAIDed HDDs have been doing that well for much lower cost. In general servers are extremely interested in random write performance and IOPS, even more than desktop users.

      • by thue (121682) on Sunday November 22, 2009 @02:50PM (#30195256) Homepage

        The slow random write will also be a problem for some very common server workloads, such as databases.

        • Re: (Score:2, Insightful)

          by mysidia (191772)

          The drive outperforms the mechanical drive in IOPs and block reads/writes which is what matters.

          Databases actually tend to use larger block reads and writes, the drive would be perfect for most databases, that is, database load is just the type of load where this drive is better than other SSDs...

          With suitable amount of system memory and host controller with reasonable cache, this drive would be phenomenal in table scan performance.

          It's application loads that are heavy in small random reads and write

          • by evanbd (210358)
            The drive to compare it to isn't a mechanical disk, but the Intel X-25M (or X-25E). The X-25M does better at small random writes, for a similar cost per GB. If you care about performance per dollar, the Intel drive still wins.
            • Re: (Score:2, Informative)

              by mysidia (191772)

              The X-25M is an 160gb HDD, half the capacity. The IOPs of SSD drives are so large, that in fact, a 30 or 40% IOPs difference is basically irrelevent for DB apps; transfer throughputs for random reads/writes at various block sizes are the most telling factor.

              Being able to quote 30000 IOPS is useless, if that number cannot be sustained with at least a 256K blocksize, commonly used for filesystems and database apps. Small random reads/writes are rare in the most demanding real-world apps.

              And the Coll

              • Re: (Score:3, Interesting)

                by AllynM (600515) *

                I think you're missing something about IOPS. With a 256K block size, you'd be lucky to crack 1000 IOPS over a SATA 3Gb/sec link. At such a large block size you hit the interface bandwidth limit way before you hit any IOPS limit.

                Multithreaded database applications do not hit a drive with sequential 256K block requests. Under load, there will be several of those requests occurring simultaneously. Given the timing, a non-NCQ drive may receive the parallel requests rapidly alternating among multiple 256K st

          • by afidel (530433)
            The only databases where tablescan performance matters is datamarts, for almost everything else it's random I/O as in most OLTP systems. Oracle and NTFS (SQL Server) both default to 4KB access units though I've gone as high as 16KB for Oracle and 64KB on NTFS (the max for current versions of NTFS)
      • by evanbd (210358)
        The 256 GB drive is cheaper -- into the price range of the dedicated power user. OCZ doesn't sell to the server market, they sell to the desktop user. The power user on a desktop machine still cares about random write performance. They're far more likely to *also* care about sequential read / write performance, but they still care about random writes too.
      • Re: (Score:2, Insightful)

        by Sulphur (1548251)

        Just like old times. Sell your car and buy a hard disk.

    • by AllynM (600515) *

      4KiB by itself is not enough, it needs to be done at different queue depths as well. I do collect data on that but it's not part of the standard set of graphs we put out. Our Workstation test gets close enough on that one, and uses a more realistic mix of reads and writes. Nobody hits their drive with 4KiB random writes all day, so I refrain from going with those specific numbers.

      Starting with the review prior to this one, I revised the Average Transaction Time graphs to show something more along the lin

      • by evanbd (210358)

        Yes, different queue depths is good too.

        No one cares only about the raw random write speed, just like they don't care only about the raw sequential speed. There are times when either matter. That's why I'd like to see both the raw numbers and the numbers for a specific workload.

        I find the average transaction time graphs hard to read. The semi-log plot format is a bit odd. It's also a bit hard to see relative changes when everything interesting is in the bottom 25-50% of the graph. Relative performance o

        • by AllynM (600515) *

          > I find the average transaction time graphs hard to read. The semi-log plot format is a bit odd. It's also a bit hard to see relative changes when everything interesting is in the bottom 25-50% of the graph. Relative performance of two drives is easy enough to see, but between the semi-log plot and the sparse grid lines, it's really hard to see the scaling effects you're talking about.

          Agreed, and I thank you for the feedback. The data itself is logarithmic, not the scale. Highly optimized drives produ

          • by evanbd (210358)

            The plot was semi-log: the x-axis values were 1, 2, 4, which is an exponential progression, but were plotted at equal spacing. You would make it a linear plot by having the spacing between the depth=1 and depth=2 points be half the space between the depth=2 and depth=4 points. More interesting to me, I suspect, would be the log-log plot where the y-axis also varies logarithmically; then the linear behavior (doubling of queue depth causes doubling of average time) would show as linear, but so would other c

            • by AllynM (600515) *

              Righto - I had Y-axis on the brain. The typical configuration for IOMeter has it ramp queue depth logarithmically. I could shift the axis to linear, but there would be missing data points. Adding tests to fill in those points adds greater risk of fragmenting the drive during the test.

              Excellent observation on the file copy test. I'll take that on board as well.

              evanbd - if you see this, lets continue via email.

              Allyn Malventano
              Storage Editor, PC Perspective

  • by thomasdz (178114) on Sunday November 22, 2009 @02:39PM (#30195168)

    People think all SSDs are the same. They aren't. Consumer SSDs are typically MLC and have a failure rate far above "enterprise" SSDs which are SLC. I wish you could buy consumer SLC SSDs

  • Useless (Score:5, Insightful)

    by ShooterNeo (555040) on Sunday November 22, 2009 @02:55PM (#30195288)

    This kludgey design is a bad idea for several reasons :

    1. Despite throwing the kitchen sink at the problem, those indilinx chips are still much slower than Intel's controller at small, random reads and writes.

    2. Since the drive needs four indilinx controllers rather than 1, some complex packaging, AND 3 RAID controllers it's going to cost a lot more per gigabyte. It's probably also more failure prone. And the MSRPs bear that out : this is a lot more expensive than the MSRPs for the equivalent Intel product.

    3. Doesn't support native TRIM support

    4. Biggest problem of all : the drive is bandwidth starved because it's on the SATA bus rather than on the PCI express bus. Furthermore, those slow internal RAID chips don't help matters. So instead of supporting sequential reads at 600 megabytes/second, it's capped at about 240. Lame.

    • Re:Useless (Score:4, Interesting)

      by seanalltogether (1071602) on Sunday November 22, 2009 @03:45PM (#30195684)
      What I don't understand is that if they're going to make a RAID SSD in a 3.5 enclosure, why don't they give it 2 SATA links in so they can saturate 2 buses? In fact, how many SATA links could you support in a single 3.5 enclosure?
      • by MrNemesis (587188)

        Because if you give it 1+n SATA controllers, you're essentially handing the "RAID" part of the setup to some software drivers - either host-based RAID (which is notoriously slow with SSD's) or some SSD-specific code that does... host based RAID, although hopefully more reliably. This is the only way you can get more bandwidth out of SATA2 connections - by glomming them together. SAS is a different beast (we have a bunch of 6Gbps SAS caddies at work) but I didn't see anywhere that the SSD supports both SAS a

  • And (Score:4, Informative)

    by DaMattster (977781) on Sunday November 22, 2009 @03:19PM (#30195458)
    It comes with a Collosus of a price tag :)
  • Forget that (Score:3, Insightful)

    by Yvan256 (722131) on Sunday November 22, 2009 @03:24PM (#30195498) Homepage Journal

    I'm waiting for the Guardian model.

    • by Jesus_666 (702802)
      Don't put both in the same computer, though. Otherwise you'll be swamped with "THERE IS ANOTHER DRIVE" messages.
  • by haruchai (17472) on Sunday November 22, 2009 @03:53PM (#30195738)

    and the companies ( Hello, Samsung!) should be ashamed. It wasn't until a few years ago that MLC was commercially viable but it only increases
    by a factor of TWO. That's one of the lowest, most pointless tradeoffs ever in recent computing.

    So, I get merely TWICE the storage for a TEN TIMES reduction in average component life, a 40% reduction in write speed, without fancy controller
    redesign, and we get to enjoy all the ludicrous "benefits" of MLC for the price that SLC would have been anyway, through market forces and silicon die shrink

    • Re: (Score:3, Informative)

      by Anonymous Coward

      MLC echos a design philosophy in computer engineering these days. Build stuff to work until exactly one day after the warranty expires. Even with adding myraid ways of error correction, it is like substituting road apples for apples for Thanksgiving pie, and pouring on the spices and sugar to minimize the poo taste.

      MLC just needs to be shitcanned and the focus be on getting SLC technology better/faster/cheaper.

      • by haruchai (17472)

        I agree with you about MLC but I can't see how it'll happen. There are only a handful of manufacturers and the average user has no idea of the differences, advantages and tradeoffs so unless some other tech comes along to shake things up, we're stuck with the status quo.

      • MLC just needs to be shitcanned and the focus be on getting SLC technology better/faster/cheaper.

        Keep in mind that no matter how good SLC gets, MLC will cost about half or a third (for the 3 bits/cell version that I hear is in the works) as much (or less really, due to higher volumes).

        • Keep in mind that no matter how good SLC gets, MLC will cost about half or a third (for the 3 bits/cell version that I hear is in the works) as much (or less really, due to higher volumes).

          When I looked last month, SLC is 5x to 8x as expensive as MLC in $/GB.
    • Re: (Score:3, Interesting)

      by RAMMS+EIN (578166)

      Are MLCs really that bad? I'm finding budget SSDs with MLCs, with warranties of several years. If the drive lasts several years, I'm a satisfied customer. I don't care if an SLC-based drive would have lasted longer, because by that time, I'll be replacing my drive with a larger one, anyway.

    • and the companies ( Hello, Samsung!) should be ashamed. It wasn't until a few years ago that MLC was commercially viable but it only increases by a factor of TWO. That's one of the lowest, most pointless tradeoffs ever in recent computing.

      So, I get merely TWICE the storage for a TEN TIMES reduction in average component life, a 40% reduction in write speed, without fancy controller redesign, and we get to enjoy all the ludicrous "benefits" of MLC for the price that SLC would have been anyway, through market forces and silicon die shrink

      Right then, that just means that everyone will buy SLC drives and the MLC ones will be a complete flop. Right?

      Or maybe, just maybe, $$/GB actually matters to people, nobody cares about expected lifespans of more than 7 years because they'll have a new computer with a 10x bigger drive by then, and even with the slower writes they're still fast enough to generally put things back to being CPU/memory bound.

      • by haruchai (17472)

        The transition was invisible to most - and that was the problem. It's not as if there were companies pitched into either the SLC or MLC camp.

        I only became aware of the transition back when USB flash drives were getting above 4 GB and found that finding ones >4GB with really fast writes speed was nearly impossible. I then stumbled onto a PDF from SuperTalent explaining the differences.

        Of course, $$/GB matters in the consumer market. Lower cost is what drives it, but the SLC drive pricing doesn't reflect t

  • HDD is cheap storage nowadays, but SSD with ist speed is great for putting programs on it. You rarely need more than 16 Gb for your installed software.
    Recently, I have bought a Verbatim SSD 16Gb Expresscard for my laptop, made it EXT2 and copied my software on it (left /var /tmp and such on HDD to avoid writes) and got my 25-seconds boot. And by that I mean complete system start with KDE4 and several apps like kopete, kbluetooth, knetworkmanager, klipper, korganizer + some plasmoids on desktop. Prior to SSD

Given its constituency, the only thing I expect to be "open" about [the Open Software Foundation] is its mouth. -- John Gilmore

Working...