Slashdot Mirror


Itty Bitty SCSI Hard Drive Arrives

Bender writes "The Tech Report has a review of the new Seagate Savvio hard drive. This little SCSI drive is roughly one-third the size of the Cheetah 10K-RPM drives so popular for servers, but the benchmarks all show it performing about the same. Not only that, but noise levels and power consumption are both lower than 3.5" SCSI drives. Is it time for 1U servers to convert to 2.5" hard drives?"

40 of 266 comments (clear)

  1. It has to be said.... by thewiz · · Score: 3, Funny

    "Phenomenal H4x0r powers; itty bitty living space!"

    --
    If "disco" means "I learn" in Latin, does "discothèque" mean "I learn technology"?
  2. 1U Servers To 2.5" Drives by TheFlyingGoat · · Score: 5, Interesting

    The IBM 336 servers that just came out use the new 2.5" SCSI drives. Instead of being able to fit 2 drives, they can fit 4. It's pretty cool stuff. The drives were slightly more expensive, but it was well worth it to us.

    --
    You have enemies? Good. That means you've stood up for something, sometime in your life. --Winston Churchill
    1. Re:1U Servers To 2.5" Drives by Anonymous Coward · · Score: 3, Insightful

      How much is slightly? The article says that these drives cost 3 times as much per drive ($447 vs $150).

    2. Re:1U Servers To 2.5" Drives by TheFlyingGoat · · Score: 4, Informative

      We get good government discounts so we paid about 50% more on each 146GB 10K drive. That's negligible when the total cost of the server is over $10K and a decent fibre NAS runs > $20K.

      --
      You have enemies? Good. That means you've stood up for something, sometime in your life. --Winston Churchill
    3. Re:1U Servers To 2.5" Drives by tmasssey · · Score: 5, Informative
      I just read about those machines about 2 weeks ago on IBM's site. When I saw them I thought, "Oh crap: IDE notebook drives with a SCSI chip stuck on them in a server. What *were* they thinking?!?"

      I must say, though, now having seen the tests and, more importantly, the photographs, that those look *nothing* like a notebook (IDE) hard drive, with their aluminum foil-quality shell almost no real structure. They look like 3.5" hard drives scaled down: still rugged, just small.

      I say bring it on! Of course, given my (and my client's) needs, I don't buy rackmount servers... :)

    4. Re:1U Servers To 2.5" Drives by geoffspear · · Score: 3, Insightful

      Wait a second... why would you want a more rugged casing on a drive that's in a machine bolted to a rack in a machine room somewhere than you do in a machine you're walking around with all day? Isn't that a bit counterintuitive?

      --
      Don't blame me; I'm never given mod points.
    5. Re:1U Servers To 2.5" Drives by tmasssey · · Score: 4, Insightful
      No. I don't want a part that's more rugged for my servers. I want a more rugged part. Period.

      However, don't forget that the biggest reason notebook hard drives are not more solidly built is because of weight, not size. When every notebook builder is struggling to gain fractions of *ounces*, every bit of extra steel on a hard drive counts. Hence, the cheap, flimsy structure.

      Have you ever seen a notebook hard drive? All of the ones I've seen in the last three years have a warning on them: do not push on drive! The top of the drive is little more than stiff foil. If you push on it, it will break the drive.

      So, no, I do not want a part specifically engineered to be as thin and flimsy as possible in my server. I don't really want them in my notebook, either, but I don't have a choice there...

    6. Re:1U Servers To 2.5" Drives by steve_l · · Score: 4, Informative

      Having been involved with notebook engineering in the past, I can assure you that the notebook drives -esp. the hitachi ones- are well engineered for the weight.

      The biggest issue with causes of failures is how well they are mounted, and that is where different ODM designs can vary wildly; or even the same ODM design with a different vendor's case round it.
      Some drives were only mounted on one side, so every shock got amplified. Others were in "quick swap on failure" units that almost guaranteed failure, they were so unsupportive of shock. Same goes for hot swap CD/DVD drive trays, BTW.

      The emergence of "Consumer grade" laptops has actually done a lot to improve the Annualized Failure Rate (AFR). These ones dont have so many hotswap options, but instead can lock down everything to be sure it stays supportive.

      We have also done tests shipping packaged systems around by fedex with a logging accelerometer in place of an HDD. you get some interesting figures, but all well within the safety range of things.

      One tip though, always tuck the laptop in behind a seat safely before you go driving down windy back roads doing italian-style "optimistic overtakes". Some things are way outside the envelope.

  3. Perfect... by Supp0rtLinux · · Score: 3, Interesting

    I see two awesome goodies with this:

    1) I can now fit 6 HDs in my 1U server instead of only three

    2) I can finally have SCSI performance on my laptop if I can ever get one with onboard SCSI. Of course, heat is still an issue...

    1. Re:Perfect... by Wesley+Felter · · Score: 5, Insightful

      Actually, now you can fit 10 drives in 1U instead of 4.

      SCSI in laptops? Keep dreaming.

    2. Re:Perfect... by Dr_Makarov · · Score: 4, Interesting

      Grab an Apple Powerbook 100 series laptop.

    3. Re:Perfect... by jcostantino · · Score: 4, Interesting

      If I recall correctly, very early Powerbook (Duo 200 series?) had 2.5" SCSI hard drives.

      --
      Reviews with a twist! http://www.sardonicbastard.com
    4. Re:Perfect... by Tanktalus · · Score: 3, Informative

      Actually, the performance gain I saw in SCSI was simply the fact that the SCSI card took care of large amounts of my I/O processing, leaving my CPU free to do, you know, CPU stuff.

      SCSI's advantage is not solely in the performance of its devices.

      When I'm burning CDs on my IDE-based CD burner, it chews up nearly all my system resources on my puny 1.8GHz processor. But on my old 486DX2/66MHz system, with 5 SCSI disks (no RAID) and SCSI CDROM, I could have all these lit up without any drain to my system. Do I miss those days or what. <sigh>

    5. Re:Perfect... by Ford+Prefect · · Score: 5, Informative

      When I'm burning CDs on my IDE-based CD burner, it chews up nearly all my system resources on my puny 1.8GHz processor.

      Try enabling DMA and suchlike, so the IDE chipset takes care of large amounts of your I/O processing.

      Between SuSE 8.1 and 9.0, my PC's IDE chipset gained DMA support for writing CDs and stuff. The machine went from being unusable when writing a CD to taking up a few percent of processor time, on my punier 1.1GHz processor. Okay, it's not SCSI performance by any means (although it talks SCSI over the IDE bus, heh) but it's still a big improvement.

      Actually, the last SCSI device I bought new was a 230MB hard disk for my Atari ST, for a few hundred pounds. I take it things have improved since then. :-)

      --
      Tedious Bloggy Stuff - hooray?
  4. Is it time? by Roadkills-R-Us · · Score: 3, Insightful

    Or is it past time?

    Either way, it's time now. How many of these can we fit in a 1U front panel and still have room for
    air inlets at reasonable volume, lights and switches? And preferably a video connector and two USB ports?

  5. OMG Smaller Hard Disks? by Exmet+Paff+Daxx · · Score: 3, Funny

    Shocking.

    --
    If guns kill people, then CmdrTaco's keyboard misspells words.
  6. My only problem.. by dadragon · · Score: 3, Informative

    My only problem with this is that SCSI disks are far too expensive for me. I'd like to have one in my desktop, but it won't happen any time soon. I'll stick with SATA for now.

    --
    God save our Queen, and Heaven bless The Maple Leaf Forever!
  7. Interesting by MasTRE · · Score: 5, Interesting

    I think that once the trend of "bigger, faster" stops, some sanity will come to computing in general. Some applications don't need the absolutely fastest performance out there, especially when that performance comes at the price of size, power consumption and heat dissipation. Most servers would be better off with a slightly slower-performing drive that uses less power and dissipates less heat. Maybe this is the start of something beautiful ;)

    --
    Must-not-watch TV!
    1. Re:Interesting by TheFlyingGoat · · Score: 4, Interesting

      You make good points, but there's another approach for servers to take. Purchase VMWare and one big server to do the job of many individual servers. You get far less power consumption and heat, make use of most of the processing power of the server (instead of running at 10% processor most of the time), and make it far easier to upgrade (increase RAM in 1 physical machine and you increase it in all the virtual ones).

      --
      You have enemies? Good. That means you've stood up for something, sometime in your life. --Winston Churchill
    2. Re:Interesting by AKnightCowboy · · Score: 4, Informative
      That is a good point, but remember that when you install that RAM in the 1 physical machine, you have to take all of the virtual machines + the physical machine offline. Single point of failure.

      Naw, you just move the virtual machines over to secondary systems that have the spare capacity, bring down the box, upgrade it, move the virtual machines back onto it, all without shutting anything down. See VMWare ESX, Vmotion and VirtualCenter for details on their site. Course, to take advantage of moving the machines without shutting them down you need a SAN on the backend. It's mainframes all over again.

    3. Re:Interesting by gbjbaanb · · Score: 3, Informative

      Yes: check out RackForce

      They do 'DDS' servers under several plans, the cheaper the plan, the more virtual servers share the same physical hardware. At the top end, you get the whole server yourself.

      The big plus with this system is that they can migrate your server to new hardware by copying a single file or directory - and they will, downtime for server upgrades is in the matter of seconds (copy, turn off old VM, turn on new VM), and you can migrate to new plans with the same ease.

      The VM technology is SWSoft's Virtuozzo which comes with some features to prevent 1 VM from taking over the entire hardware - you can set it so each VM will be guaranteed a minimum amount of resources.

  8. Failure rate? by seagar · · Score: 5, Informative

    That's what i'm most concerned with..I have never cared much about the noise level of SCSI drives in my SERVER ROOM. It's supposed to be loud in there. Lower power consumption is a plus.

    Back to failure rates, I have noticed a slip in the quality of my Seagate drives lately (IDE, SATA, and SCSI). They just seem to fail more often than they used to. I used to brag about how rock solid my Seagates were. However, I also seem to remember Seagate extending their warranty coverage to something like 5 years? Maybe this is a sign that I just had bad luck with my drives..it's been known to happen.

    --

    home of the original cupholder
    1. Re:Failure rate? by Jucius+Maximus · · Score: 5, Informative
      " Yea, I'm sure their drives have an average lifespan of 159 years. Next caller."

      This is a common misconception. The MTBF refers to the time before a failure in group of drives. So if you have 120 of these 1,400,000 MTBF drives in your server room, then you can expect to go 1 year, 121 days (on average) between replacing drives. That should help you plan your IT spending budget too.

      Or perhaps a company deployed 5000 laptops, all with these drives in them: You can expect to go about 12 days between failures. Back up your drives, people! Even with SCSI-level MTBF numbers, statistically failures are not all that uncommon.

  9. Dr. Evil by kettleoffish · · Score: 4, Funny

    And I shall call it... MINI-SCSI!

  10. specialized boot drives by TheSHAD0W · · Score: 4, Interesting

    Sounds like someone should be making specialized boot drives, 1.5" or smaller, with 5 gig capacities and super-fast seek times and rotation rates. The smaller the platter diameter, the less strain on the bearings and the more reliable they'll be at ludicrous speed.

    1. Re:specialized boot drives by synthparadox · · Score: 3, Informative

      The infamous RocketDrive RocketDrive.

      Or you can use Flash drives, less expensive but a bit slower.

  11. Re:Is it really an upgrade? by Kenja · · Score: 4, Informative
    "...why would server owners want to "upgrade" to a smaller, quieter, more expensive drive if they're not even going to get a performance increase?"

    More drives equals more performance. A six drive RAID-5 will outperform a three drive RAID-5. With smaller drives you can fit more of them in a 1U system.

    --

    "Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
  12. Density by ZxCv · · Score: 4, Informative

    Correct me if I'm wrong, but why would server owners want to "upgrade" to a smaller, quieter, more expensive drive if they're not even going to get a performance increase?

    Perhaps they might see the value in fitting more drives into the same server enclosure?

    --

    Perl - $Just @when->$you ${thought} s/yn/tax/ &couldn\'t %get $worse;
  13. It's time to go 2.5" by egarland · · Score: 5, Informative

    The transition to 2.5" drives should begin now. The 1U server market would be a great place to start because space, airflow and power utilization are all problems with 3.5" drives in 1U servers. History tells us within a few years most drives will probably be 2.5". We are at the point where the 2.5" drives are fast enough and have enough capacity to be appropriate for the common desktop user as well as the high end server user. The price premium is currently too high for wide spread desktop adoption but that's less of an issue in the server realm.

    The material, storage and transportation costs of 2.5" drives are all dramatically lower than 3.5" so in the end, they should become cheaper than 3.5" drives as the technology ages. Since laptop sales are so high the economies of scale for 2.5" drives are there. All we need now is for a company to streamline their manufacturing to bring the cost down to the levels of 3.5" drives and the en-mass transition will begin.

    I for one, can't wait to have 8 drive raid array that fits in two 5.25" drive bays.

    --
    set softtabstop=4 shiftwidth=4 expandtab nocp worlddomination
  14. Small SCSI drives are nothing new by danuary · · Score: 4, Interesting
    ...In fact, they're something old -- Macs once upon a time used laptop-sized SCSI disks; so did Sun's SPARCstation Voyager. In the case of the Voyager, a few were made with a 1GB 2.5" (laptop form factor) SCSI disk (the rest had 340MB and 520MB).

    I think the push for IDE came around this time and the market died for 2.5" form factor SCSI. Nice to see it's being revived.

    Wish I still had my trusty old Voyager - because it'd be fun to see if I could get one of these newfangled drives working in it with some sort of an adapter!

  15. 160 years MTBF by kclittle · · Score: 5, Funny
    1.4e6/(24*365.24) = 159.71 years, to be picky about it. I see these figures on modern drives and, frankly, I don't believe it. But, that doesn't keep me from drooling over them (which would proably shorten the MTBF, yes? :-)

    --
    Generally, bash is superior to python in those environments where python is not installed.
  16. Not surprizing by ^BR · · Score: 4, Informative

    The platter diameter in fast rotating disks have been smaller and smaller (thus explaining the not so great capacity compared to ATA drive that use full 3"5 platers, not rotating fast).

    The common platter size went from 3"5 to 3" to 2"6 to 1"8, it was only a matter of time that they decided to package it in a smaller enclosure, the 1U market explains a lot... See that very old review (Y2K) or that Seagate whitepaper (pdf) about why smaller is faster...

  17. As An Engineer Who Has... by John_Booty · · Score: 4, Funny

    ...worked in the industry for a long time, let me just give my technical impression on these 2.5" drives.

    THEY'RE SO CUTE I WANT ONE

    --

    OtakuBooty.com: Smart, funny, sexy nerds.
  18. teeny by jav1231 · · Score: 3, Funny

    Must have the PreSCSIous!

  19. Lower Power? by TFloore · · Score: 4, Insightful

    Kind of... but only on a per-drive basis.

    The article talks about putting 3U of 140GB 3.5" HD RAID storage in 2U of 73GB 2.5" HD RAID storage now, for the same total HD space for the array.

    Same storage space. Twice the number of drives. 2/3 the rack space. 44% power use PER DRIVE. That works out to 92% of the power of a 3U RAID stack, in a 2U RAID stack. Which means you just UPPED the power requirement for a fully-populated rack by about 40%.

    Congratulations, your lower power device has you using more power. And therefore dissipating more heat in the same volume. Of course, you DO get a 50% increase in storage capacity for that.

    But you still upped your total power per rack by 40% if you do that.

    Remember your ear protection. The drives are quiet, but that many fans make a lot of noise.

    --
    This is my sig. There are many like it but this one is... Oops. Frank, I've got your sig again! Where's mine?
  20. Does this mean SCSI for laptops? by steve426f · · Score: 4, Interesting
    Perfect for the Dell Precision M60 (mobile workstation)! Since I'm a non-gamer, hard drives seem to be the main bottleneck for laptops. Most of the high-end laptops last for only an hour or less on battery, users are already accustomed to using A/C, therefore battery isn't an issue for those who want SCSI performance.

    My laptop is used more as a portable workstation. PDA's are for battery powered portability!

  21. Seagate history by magarity · · Score: 4, Insightful

    Once upon a time, Seagate listened to its server customers and continued to produce and develop 5.25 inch drives while ignoring the new 3.5 inch format. KaLock and Quantum jumped into the marketplace with 3.5 inch drives and sold them to desktop makers. Seagate lost a LOT of market share by ignoring the push to smaller drives. It seems they are being proactive in moving to 2.5 inch format early before one of the other manufacturers get the jump on them this time.

  22. government pricing by magarity · · Score: 4, Funny

    We get good government discounts so we paid about 50% more on each 146GB 10K drive

    Typical...

  23. Re:This is great by slittle · · Score: 3, Insightful
    Performance the same, lower noise and power consumption--this is all great, but the most important question is: is it equally robust as the full-sized version?
    Smaller size (== less mass) and lower power (== less heat) can only be a Good Thing for wear and tear. Unless you drop it I suppose.

    because anything that is smaller is inevitably easier to scratch, as any given scratch is relatively larger
    Only if the density is higher. You'll note SCSI drives tend to lag a bit behind IDE for capacity, and 2.5" drives lag behind a LOT. Not that it matters, any scratch of note means a dead drive. The days when one could live with a bad sector or three are long gone. Once SMART reports your drive is using its spare sectors, it's time to place an order for a replacement.

    namely laptops might finally get SCSI drives to achieve much better quality and throughput than the legacy IDE we are usually left with now
    Market separation. There's nothing stopping mfr's from making high quality, high speed IDE drives. They just don't want to. If SCSI hits mainstream there will be pressure to lower the cost of SCSI, which will fuck up their profit models. Right now, if you're serious about storage, you bend over and take it with a smile as you have no other choice. SCSI-on-the-desktop/laptop gives SCSI users a choice.

    I for one quite like the SCSI zealots subsidising my cheap 256GB IDE drives, thank you :)
    --
    Opportunity knocks. Karma hunts you down.
  24. SCSI Event Request Reordering gives Speed by billstewart · · Score: 3, Interesting
    One reason SCSI disks are faster is that the smart controllers get to optimize requests after they've been sent to the disk controller. Typical tricks include reordering requests to take advantage of disk postition information that the CPU doesn't know about. I haven't benchmarked this stuff in years, but basically anybody who does ends up raving about SCSI performance.

    Another reason that SCSI disks are often faster is that they often have higher RPMs. That's not because the controller makes the disk spin faster - AFAIK it's just because the disks that spin faster are usually sold to people who want maximum performance and are willing to pay for it, so they usually want SCSI controllers.

    More spindles is obviously a Good Thing too, but that's not what makes SCSI fast. It would seem obvious that SCSI lets you support more spindles, so that would give you some speed advantages, but most SCSI disks seem to be smaller, so for any given capacity you often need more disks if you're using SCSI.

    --

    Bill Stewart
    New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks