Results 1 to 19 of 19
  1. #1
    Join Date
    Mar 2009
    Posts
    2,735

    SATA 3.5" vs 2.5"

    Hi,

    for sata two,500GB,

    3.5" vs 2.5",both 7200rpm,

    will they have the same performance and reliability ?



    thanx

  2. #2
    tomshardware.com has benchmarks for this:

    2.5", database benchmark:

    http://www.tomshardware.com/charts/2...7.27,1130.html

    3.5", database benchmark:

    http://www.tomshardware.com/charts/2...7.27,1034.html

    the top performing 500gb 3.5" drive clocks in at 104 i/o/s

    the top performing 500gb 2.5" drive clocks in at 140 i/o/s

    Of course, the 1tb caviar black 3.5" clocks in at 160 i/o/s and probably isn't much more expensive than the 500gb 2.5" scorpio black (the highest performing 2.5" 500gb drive)

    So overall I would say, if you're looking at the raw performance per drive, they're fairly similar if you make sure to buy the right models. The raw performance per dollar spent is also fairly similar. But obviously, the storage space per dollar is better in 3.5", and the power usage needed to get to a given level of performance is better in 2.5".

    You can also pack more 2.5" drives into a case, but if you're dealing with the 500gb drives, once you've got more drives in a server than you have onboard sata ports for, the cost of addon sata cards, and the cost of a case that supports more than a half dozen drives, is going to be more expensive than the cost of the extra drives themselves.

    And although the power *use* savings are significant, the power *cost* savings really aren't. Even if you save 5 watts per drive, a 20 drive server would only save 1a total. Easy enough to get a 40a rack (32a usable) for $1000 give or take, putting your cost of 1a of power at around $31 / mo. Saving $31 / mo spread out over 20 hard drives, that's what, $1.50 / drive / mo? Not totally insignificant, but not particularly a big deal either.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  3. #3
    Join Date
    Apr 2003
    Location
    San Jose, CA.
    Posts
    1,622
    Reliability more or less the same...
    But, on Performance 2.5" wins.
    Daved @ Lightwave Networking, LLC.
    AS1426 https:/www.lightwave.net
    Primary Bandwidth: EGIHosting (NLayer, NTT, HE, Cogent)
    Xen PV VPS Hosting

  4. #4
    Join Date
    Nov 2009
    Location
    Cincinnati
    Posts
    1,583
    We have a san with 24x 300gb 10k 2.5" and 12x 300gb 15k 3.5" disk, the 10k 2.5s are as fast if not FASTER than the 3.5" 15k.

    I can't wait for 600gb 15k 2.5" disk, those will be sick.
    'Ripcord'ing is the only way!

  5. #5
    Quote Originally Posted by Visbits View Post
    We have a san with 24x 300gb 10k 2.5" and 12x 300gb 15k 3.5" disk, the 10k 2.5s are as fast if not FASTER than the 3.5" 15k.

    I can't wait for 600gb 15k 2.5" disk, those will be sick.
    At the price that costs... I'm less enthused.

    A 147gb 15K SAS costs around the same price as a 120GB intel X-25M SSD, and the SSD is going to blow away the SAS drive in performance and power use. I see SAS becoming more and more marginalized over time, although SATA will have some room to run for a few more years due to the massive storage space and low cost.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  6. #6
    Join Date
    Mar 2010
    Location
    Houston, TX
    Posts
    164
    I'm glad I stumbled upon this thread. I never realized that 2.5" drives could have more performance than 3.5" drives with the same speeds. It makes sense because of its smaller physical size but I just had not given it any thought.

    The one other thing I would say to keep in mind is not having drives with dual platters such as some 2.5" 640gb units. From my personal experience I can tell that they slow down tremendously in performance as they fill up data across the two platters.
    Andy Salazar - Steadfast Networks: Cloud, Dedicated and Colocation
    Email: andy.salazar @steadfast.net - Sales/Support: 312-602-2689
    DC Locations: 350 E Cermak, 725 S Wells Chicago IL and Edison NJ

  7. Quote Originally Posted by funkywizard View Post
    At the price that costs... I'm less enthused.

    A 147gb 15K SAS costs around the same price as a 120GB intel X-25M SSD, and the SSD is going to blow away the SAS drive in performance and power use. I see SAS becoming more and more marginalized over time, although SATA will have some room to run for a few more years due to the massive storage space and low cost.
    however, lots of WHT members who have extensive SSD experiences will tell you that any MLC based SSD's are not intended for enterprise use, and they can get "chewed" up in short few months in production environment.

    bear in mind that 2.5" SAS and VelociRaptors are built for enterprise/disk array use, but most 2.5" SATA drives are not.

  8. #8
    Quote Originally Posted by Andy Salazar View Post
    I'm glad I stumbled upon this thread. I never realized that 2.5" drives could have more performance than 3.5" drives with the same speeds. It makes sense because of its smaller physical size but I just had not given it any thought.

    The one other thing I would say to keep in mind is not having drives with dual platters such as some 2.5" 640gb units. From my personal experience I can tell that they slow down tremendously in performance as they fill up data across the two platters.
    I think you misunderstand the platters.

    Hard drives fill data in from the outside of the drive inwards. The outside of the drive spins at a higher linear speed, since the RPMs are the same but that portion of the disk is physically larger. When writing to platters, it writes to both platters at once, it doesn't start on one, fill it up, and move to the other.

    No matter what kind of drive you have, it will work slower towards the inside of the disk (end of the drive) than towards the outside. This problem is even larger in 2.5" drives, since the inner portion of the disk is smaller than on 3.5" drives.

    Because the 2.5" drives are smaller, seeking is faster, it doesn't take as long to go from one part of the disk to the other, compared to a 3.5" drive. However, because the disk is smaller, sequential reads / writes are slower. Typically, in a server environment, sequential access speeds are not your bottleneck, seeks are, so 2.5" can be a bit faster, but usually is pretty close to the same speed.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  9. #9
    Quote Originally Posted by [email protected] View Post
    however, lots of WHT members who have extensive SSD experiences will tell you that any MLC based SSD's are not intended for enterprise use, and they can get "chewed" up in short few months in production environment.

    bear in mind that 2.5" SAS and VelociRaptors are built for enterprise/disk array use, but most 2.5" SATA drives are not.
    I would take an SSD over a raptor any day. I've had failure rates on raptors of around 30% using a dozen drives over a couple years.

    SAS, sure, those are well known to have low failure rates. MLC SSD might not have proven itself yet as a replacement for SAS, but with the prices being so similar and the performance being so much better, it's only a matter of time before they can either be improved (if needed), or their track record established.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  10. #10
    Quote Originally Posted by Andy Salazar View Post
    I'm glad I stumbled upon this thread. I never realized that 2.5" drives could have more performance than 3.5" drives with the same speeds. It makes sense because of its smaller physical size but I just had not given it any thought.

    The one other thing I would say to keep in mind is not having drives with dual platters such as some 2.5" 640gb units. From my personal experience I can tell that they slow down tremendously in performance as they fill up data across the two platters.
    yeah, it was strange for me to wrap my head around this originally as well.. but, its a physics thing.. smaller circle - so, if all else is equal, and the "circle" spins at the same speed - the smaller circle will be faster then the larger circle. As someone else noted, often times, a smaller circle spinning more slowly (ie 10k) is often faster then a larger circle spinning 15k... Of course solid state drives (SSD) changes this relationship - but, for traditional drives, smaller is generally better

  11. #11
    Quote Originally Posted by cartika-andrew View Post
    yeah, it was strange for me to wrap my head around this originally as well.. but, its a physics thing.. smaller circle - so, if all else is equal, and the "circle" spins at the same speed - the smaller circle will be faster then the larger circle. As someone else noted, often times, a smaller circle spinning more slowly (ie 10k) is often faster then a larger circle spinning 15k... Of course solid state drives (SSD) changes this relationship - but, for traditional drives, smaller is generally better
    That's actually completely backwards I have to say.

    The smaller circle is not faster, as the linear speed is lower at a given RPM. The difference is, a head moving from the outside of the disk to the inside of the disk doesn't have as far to travel, and so moving from one part of the disk to the other is faster.

    You can achieve the same thing by only formatting a 3.5" drive with half of the drive's capacity, as all the data will be stored closer together (this is known as short stroking)
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  12. #12
    Quote Originally Posted by funkywizard View Post
    That's actually completely backwards I have to say.

    The smaller circle is not faster, as the linear speed is lower at a given RPM. The difference is, a head moving from the outside of the disk to the inside of the disk doesn't have as far to travel, and so moving from one part of the disk to the other is faster.

    You can achieve the same thing by only formatting a 3.5" drive with half of the drive's capacity, as all the data will be stored closer together (this is known as short stroking)
    we said the same thing man - though, I guess you worded it better.. but, yeah, effectively, the smaller circle is faster at the same speed (which is what I said) - and obviously it is because of distance... SSD's, as I said, are changing this..

    Interesting re 3.5" - but, whats the point of formatting to half the volume. You would be better off just buying smaller 2.5" drives..

  13. #13
    Quote Originally Posted by cartika-andrew View Post
    we said the same thing man - though, I guess you worded it better.. but, yeah, effectively, the smaller circle is faster at the same speed (which is what I said) - and obviously it is because of distance... SSD's, as I said, are changing this..

    Interesting re 3.5" - but, whats the point of formatting to half the volume. You would be better off just buying smaller 2.5" drives..
    I think I'm still misunderstanding what you're saying. If you mean "faster" in terms of, better seeking performance, then yes, it could be. But it's pretty close really, and I wouldn't expect much better performance at the same RPM.

    If you mean faster, in terms of, you can read data off the disk faster, or you mean that the data flies past the read / write head faster, then no, the 2.5" is demonstrably slower in those regards.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  14. #14
    Quote Originally Posted by funkywizard View Post
    I think I'm still misunderstanding what you're saying. If you mean "faster" in terms of, better seeking performance, then yes, it could be. But it's pretty close really, and I wouldn't expect much better performance at the same RPM.

    If you mean faster, in terms of, you can read data off the disk faster, or you mean that the data flies past the read / write head faster, then no, the 2.5" is demonstrably slower in those regards.
    sorry, but, what are you getting at exactly? I made some pretty general statements - and I stand by them

    what I said originally is this:

    yeah, it was strange for me to wrap my head around this originally as well..
    and I noticed the same overall performance concepts as this poster mentioned

    We have a san with 24x 300gb 10k 2.5" and 12x 300gb 15k 3.5" disk, the 10k 2.5s are as fast if not FASTER than the 3.5" 15k.
    and we have noticed this over 100s of examples within our NOCs
    thanks

  15. #15
    Join Date
    Nov 2009
    Location
    Cincinnati
    Posts
    1,583
    Quote Originally Posted by funkywizard View Post
    At the price that costs... I'm less enthused.

    A 147gb 15K SAS costs around the same price as a 120GB intel X-25M SSD, and the SSD is going to blow away the SAS drive in performance and power use. I see SAS becoming more and more marginalized over time, although SATA will have some room to run for a few more years due to the massive storage space and low cost.
    Yeah and will die after 1/1000th the write cycles... no thanks bro.

    X-25E are the only SSD worth my time and their price does not justify their performance in my world.

    I need capacity to utilize performance, can't utilize performance if the capacity isn't available... cheaper to stack on disk.
    'Ripcord'ing is the only way!

  16. #16
    Quote Originally Posted by Visbits View Post
    Yeah and will die after 1/1000th the write cycles... no thanks bro.

    X-25E are the only SSD worth my time and their price does not justify their performance in my world.

    I need capacity to utilize performance, can't utilize performance if the capacity isn't available... cheaper to stack on disk.
    If you do the math, and look at the number of writes per second a SAS drive can do, and divide that into the number of writes you can do to an intel MLC before it fails, you'll see that, it would last several years writing at the same level that a SAS drive is capable of. It's like if you had a car that went 20 miles an hour, and a car that went 200 miles an hour, and they both would last until 100,000 miles. Which car will last longer? The slower car will last "longer" if you drive both cars at maximum speed, but if they're the same price, and you drive them the same speed, that's a really irrelevant distinction.

    It might degrade too soon if you're using it to write at 10x the level of speed that a SAS drive is capable of, but to get the same level of performance you'd need 10 SAS drives, so it's really not a fair comparison.

    A $200 SSD drive that goes 10x as fast and lasts a year, vs 10x $200 SAS drives to match the same speed, but last until they break.

    Or a $200 SSD drive that goes 1x as fast and lasts 10 years, vs 1x $200 SAS that goes 1x as fast and last until it breaks.

    And if you look at the numbers, the SLC is not that much better. Sure, each cell can be written to 10x as many times before it dies, but you have a lot fewer cells for the same money. In terms of write cycles before death divided by cost, the SLC is only about twice as good as the MLC.

    In any case, if you've got a read-heavy workload, I just don't see any reason at all to use SAS over SSD with the 120gb intel down to $200 vs the same price for a 147gb SAS drive. If you have a mix of reads and writes, I'd still use the SSD. If you were just doing tons of writes constantly, like, say, for a video surveillance system, sure, SAS is a much better choice.

    The situations and conditions where SAS is a better option than SSD are limited and growing smaller every day, both in terms of price, performance, and storage. Sure, you may want to use SAS *today* for some legitimate reasons in some narrow circumstances, but my original point was that I'm hardly going to get excited about the *future* of SAS, like you had originally pointed out (" 600gb 2.5" SAS I can't wait").

    The situations where SAS is a better choice than SATA is limited to those areas where performance matters a lot more than storage capacity. A $200 147gb SAS drive is only twice as fast as a $100 1tb SATA drive, while providing only 1/7th the disk space. 4x $100 sata drives in a raid 10 should perform similarly to 2x $200 SAS drives in Raid 1, so I don't see that as very compelling there. Now, if you're looking for an 8 drive array of SAS drives, sure, it's easier to build a server with 8 drives rather than 16 drives, I will give you that. Or even if you just don't care about disk space *at all*, you could get a 74gb SAS drive for a bit over $100, making SAS just slightly better than SATA in terms of performance per dollar. Don't even get me started on 600GB SAS drives. Why would you buy a $400+ hard drive that's only twice as fast as a $100 hard drive, when the $100 hard drive had twice as much disk space? Just get 2x $100 hard drives and call it a day, it's not even close.

    So to recap, SAS is only a tiny bit better than SATA in terms of performance per dollar, and only in some limited cases. SAS is only better than MLC SSD in the case that you're doing massive sequential writes 24/7. For a mixed workload of reads and writes, MLC blows away SAS (in my opinion, although this particular point is debateable), and for a workload that's mostly reads, there's no point in even considering SAS (a point I don't even consider debateable). Performance per dollar, performance per drive, storage per dollar, storage per drive, none of these metrics show SAS any better than your other options in any situation.

    The only last bit left for the business case for SAS vs anything else, is that the failure rate on SAS is lower than on SATA, according to some pretty good studies by people like netapp. The failure rates for SATA is low enough that anyone with a good backup plan and raid level should be just fine, but for those who want that additional assurance, SAS is more reliable, in general, today, than your other options.
    Last edited by funkywizard; 02-14-2011 at 07:58 PM.
    Phoenix Dedicated Servers -- IOFLOOD.com
    Email: sales [at] ioflood.com
    Skype: iofloodsales
    Backup Storage VPS -- 1TBVPS.com

  17. #17
    Join Date
    Nov 2009
    Location
    Cincinnati
    Posts
    1,583
    I'm not even going to read everything you wrote... to much repeated material..

    Anyway, as you said.. if you have *read-heavy* workload then an SSD is good.. well to bad most *workload* is writes and changes for database. If you have an intense read workload a ton of ram and some good caching is the best thing you can do.

    Nuff Said.
    'Ripcord'ing is the only way!

  18. #18
    Join Date
    Mar 2009
    Posts
    2,735
    how important the hdd cache is ? will 16MB/32MB/64MB have much difference?

  19. #19
    Join Date
    Nov 2009
    Location
    Cincinnati
    Posts
    1,583
    Slower drives have larger cache to make up for it.

    If you have a raid controller that has write cache the performance improvements of disk cache are minimal.
    'Ripcord'ing is the only way!

Similar Threads

  1. How efficient are topic titles like "BLOWOUT", "SPECIAL" and "KILLER OFFER"?
    By BluewaveHosted in forum Running a Web Hosting Business
    Replies: 18
    Last Post: 06-30-2009, 11:21 AM
  2. what difference between "desktop"、"work station"、"server" board ?
    By joelin in forum Colocation and Data Centers
    Replies: 1
    Last Post: 06-09-2008, 12:01 PM
  3. Replies: 39
    Last Post: 02-22-2008, 03:22 AM
  4. <select name=\"prod_id\" onchange=\"this.form.submit();\">";
    By stephenvs in forum Programming Discussion
    Replies: 4
    Last Post: 05-27-2004, 12:59 PM
  5. DDOS: GET / HTTP/1.1" 302 223 "-" "-" 800k times a day :(
    By freeflight2 in forum Hosting Security and Technology
    Replies: 15
    Last Post: 01-30-2004, 11:38 AM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •