RAID O 2-10,000 RPM 74GB Raptors

Posted:
in General Discussion edited January 2014
If you don't know already, having a 10,000 RPM boot/application drive really speeds things up on the G5's.



So I took it a step further and RAID O a pair of them. The results are interesting.



My boot time stayed the same as a single 10,000 RPM drive 25 seconds (Apple to Desktop)



I only gained two seconds booting Photoshop 7.01 7 seconds fresh, 5 seconds preloaded.



Big game play improvement from 7,200 RPM to 10,000 RPM, not much with the RAID O pair.



But good lord my large read/write speeds went ballistic.



You can see the links, X-Bench scores and chart my progress here. We all know about RAID and reliability, so I'm testing that theory out.



http://g5support.com/group/viewtopic.php?t=601





I'll add more data as my experiences continues, it's only been day #1.

Comments

  • Reply 1 of 11
    hmurchisonhmurchison Posts: 12,425member
    Noise and even thermal(permitting of course) measurements would be handy.



    I'm not suprised booting and app launching didn't speed ut appreciably. However has Photoshop run now that the cache is running off of RAID 0 10k drives?



    What improvements in game play do you see. Does your framerate actually improve?



    Look forward to seeing any other data you can provide.
  • Reply 2 of 11
    sailfishsailfish Posts: 163member
    Quote:

    Originally posted by hmurchison

    Noise and even thermal(permitting of course) measurements would be handy.



    Noise? There is no constant whine, fluid bearings after all. I duplicated a 7.8GB folder in 4 minutes 15 seconds and the heads were crunching away, just noticeable, not annoying.



    ThermographX gives a cool 86º for the Drive-bay temp, I get a IR reading of 103º F on the exposed sides each.



    Quote:

    I'm not surprised booting and app launching didn't speed up appreciably.



    Yea me too, was a bit of a disappointment, guess there is a bit more going than just loading RAM. I got a real boost from 7,200 - single 10,000 RPM.



    Quote:

    However has Photoshop run now that the cache is running off of RAID 0 10k drives?



    I need a standard Action and a photograph I can test against to compare.



    Quote:

    What improvements in game play do you see. Does your framerate actually improve?



    I noticed a visual improvement from 7,200 RPM to 10,000 RPM, not with the RAIDed pair, my ATI 9600 is probablly holding things up. I need to catch a frame rate recorder for UT2004.





    Quote:

    Look forward to seeing any other data you can provide.



    It looks the speed benefits are coming from large read/writes only.



    Anyone want to link to stats I can run?
  • Reply 3 of 11
    ghost_user_nameghost_user_name Posts: 22,667member
    Quote:

    Originally posted by Sailfish

    Noise? There is no constant whine, fluid bearings after all. --snip--





    Expect to hear the "whine" increasing over time. Enjoy them now while they are quiet.





    Quote:

    Originally posted by Sailfish





    I noticed a visual improvement from 7,200 RPM to 10,000 RPM, not with the RAIDed pair, my ATI 9600 is probablly holding things up. I need to catch a frame rate recorder for UT2004.







    Besides loading maps and textures the drives will not help gaming. Your new drives will not help with raw frame rates.



    Quote:

    Originally posted by Sailfish



    It looks the speed benefits are coming from large read/writes only.




    Ahh... raid 0, almost twice the performance at one half the reliability. If I ever go raid... raid 5 for me.
  • Reply 4 of 11
    gargoylegargoyle Posts: 660member
    Unless you are happy going for 3, 4 or 5 drives then RAID 1 should be the only choice unless you really can fill the space of two drives.



    RAID 1 writing speed is a bit slower than a single disk, but you should get some very good read improvements. Assuming that the RAID under OSX is similar to Linux, the OS chooses the drive whose heads are closest to the data needing to be read, which reduces seek time which is the real killer for small files.



    Not sure how RAID 0 handles this. Note: IMO it shouldn't be called RAID 0, because there is NO redundancy.
  • Reply 5 of 11
    sailfishsailfish Posts: 163member
    No improvement in the frame rate in UT2004, game does load the maps faster, correcto!





    Now about the reliability factor..





    LaCie has that Big Extreme, which is two 500GB drives RAID O (striped) in one housing to give it speed. With a whopping 1 TB of perhaps very sensitive data, it begs the question.



    If RAID O is so unreliable why would LaCie risk their reputation (and a lawsuit) and provide such a product?



    Perhaps when folks get into the bigger arrays?





    I'm covered, I'm a recreational user, I'm retired, I have a external drive to "option boot" from and a stock 7,200 RPM internal that can be switched out.



    So I can afford to play around a bit.
  • Reply 6 of 11
    hmurchisonhmurchison Posts: 12,425member
    RAID 0 is fine. Personally I wouldn't use it for data that I plan to leave stored for a while but for tossing data on the disc and then pulling it off I'm totally cool with that.
  • Reply 7 of 11
    sailfishsailfish Posts: 163member
    Basically that's what I'm doing, I have a external clone which I backup files to once a day.



    If the RAID goes down, I can just clone back. So far day two and it's been stable.



    Was it true eariler in Mac OS 9 we were not allowed to RAID a boot volume?



    I can't remember if it was or not.
  • Reply 8 of 11
    nguyenhm16nguyenhm16 Posts: 203member
    Quote:

    Originally posted by Sailfish





    LaCie has that Big Extreme, which is two 500GB drives RAID O (striped) in one housing to give it speed. With a whopping 1 TB of perhaps very sensitive data, it begs the question.



    If RAID O is so unreliable why would LaCie risk their reputation (and a lawsuit) and provide such a product?





    Actually, from what I've read, the volumes are spanned (the first 500GB drive covers the first half of the 1TB, and so on). Anyways, if you buy a RAID0 (or spanned) product, you are assuming the risk, you should know what your doing. It's not like everytime someone so much as mentions RAID0 somebody screams about the (lack of) reliability. If you were really hard-core, you would have four disks, a RAID0+1. Fast and reliable.
  • Reply 9 of 11
    Hard core with RAID 0+1?!? pfft... what a waste.



    Hard Core is Raid 5!
  • Reply 10 of 11
    nguyenhm16nguyenhm16 Posts: 203member
    Quote:

    Originally posted by \\/\\/ickes

    Hard core with RAID 0+1?!? pfft... what a waste.



    Hard Core is Raid 5!




    RAID 5 has horrible write speeds compared to RAID 0+1... you have to do two physical writes (one for data, one for parity) for every logical write, not to mention the computation for the parity data. Thus it's good for things that don't get written very often, but for databases, etc, that are updated frequently, performance is (relatively) terrible. Every high end storage installation I've seen (big EMCs, etc.) uses some variant of RAID 0+1; the rule of thumb was, RAID 5 only if you can't afford the spindles for 0+1.
  • Reply 11 of 11
    sailfishsailfish Posts: 163member
    Quote:

    .....<snip>





    (disclaimer: I am one of the technical people responsible for the design)



    We have a family of products:

    - Big Disk; the original; has 2 drives with a controller (Ocford 922) which spans the two disks. Two disks are combined as one logical volume, the second disk accessed after the first id filled

    - Bigger Disk; announced at Macworld SFO 2004; 4 disks, also spanned

    - Big Disk Extreme; has 2 disks striped using Oxford 912 controller; offers faster performance due to hardware striping (essentially RAID 0)



    .....<snip>



    Mike

    LaCie Employee



Sign In or Register to comment.