Please report all spam threads, posts and suspicious members. We receive spam notifications and will take immediate action!
Results 1 to 9 of 9

Thread: RAID-0 performance




  1. #1
    Join Date
    Oct 2004
    Location
    Jämtland, Sweden
    Posts
    128

    Default RAID-0 performance

    I've got three SATA-2 WD 250GiB 16MiB drives in RAID-0. As far as I know, the performance should be greater, or at least equal to that of a single such drive. But it is, in fact, worse. In SiSoft Sandra I got 52 MB/s (32-52 MB/S) in drive index, random access time at 14ms. Is this performance normal, should I use some other software to test with, or is there some known limitation in the nForce 4 Ultra MCP, like in 680i (I think that was the case at least) or is it something else?

    EDIT: Also, is there some way to check the SMART data of individual drives in a RAID array?
    EDIT2: the drive index is read performance, since write performance benchmarks can only be performed on an empty drive in SiSoft sandra
    Last edited by ifander; 03-16-2007 at 12:35 AM.

    Chieftec Bravo LBX01-B-B-SL
    580 W Hiper Type-R PSU

    Gigabyte GA-K8NXP-9 S.939
    AMD Athlon 64 Venice 3000+ at ~2470 MHz
    OCZ Gold EL PC 3700 Ram, 2.5-3-3-8 at ~456 MHz,
    2x512 MiB
    Gigabyte SilentPipe X800XL, 400/490 MHz
    Auzentech Digital X-Mystique 7.1CH DDL-RCE
    Logitech Z-5500 5.1 speakers
    1 x WD Caviar 8 MiB cache 1600JS-00MHB0 160GiB SATA-2
    3 x WD Caviar SE 16 MiB cache 2500KS-00MJBQ 250GiB SATA-2



  2. #2
    Join Date
    Nov 2001
    Location
    Bendigo Australia
    Posts
    1,788

    Default Re: RAID-0 performance

    it depends a lot on the raid controller how fast it runs.

    Personally I'd recommend against raid 0 - the main feature of raid 0 is its unreliabilty - you've just increase the change of losing all your data by a factor of 3 (if any one of the HD's fail in the array - then bingo all data is gone)

    I also don't believe you get much speed improvement from a raid 0 setup, raid 1 can be faster but the main reason for using a raid setup is to protect your data

  3. #3
    Join Date
    Nov 2003
    Location
    Minnesota, United States
    Posts
    4,543

    Default Re: RAID-0 performance

    I am in strong agreement with Kheldar. "RAID" 0 greatly decreases reliability. I disagree in that striping increases unreliability exponentially, not linearly. So three striped drives is actually going to be worse than having a 300% increase in the chance of failure. You are asking for data loss.

    As for the benchmarks, there are plenty of other factors. The controller chip itself can be an issue, as well as CPU utilization and general system resources. Depending on the controller, there's a decent chance that additional CPU usage is required for striping.

    More importantly, access time can increase from RAID 0. Depending on the specific setup, it can take longer to look for a file striped across three drives than one sitting comfortably on a single drive, with all it's parts in the same physical locations.

    RAID 1, 01, or 10 are much better choices. Firstly, they're actual RAID levels, as opposed to just striping. They actually increase reliability and speed. Reliability is very important with magnetic hard disks. Any given hard drive will fail, and there's a good chance it will be within a couple years of use. As far as speed, RAID 1 can decrease access times and increase throughput on reads, but it does tend to go slower on writes. RAID 01 and 10, on the other hand, combine the best of both worlds and will increase speed and reliability -- provided you've got a decent controller and sufficient resources.

    If you really want speed improvement, 10000RPM and 15000RPM drives and/or RAID 01 or RAID 10 are the way to go. Striping a few big 7200RPM drives produces negligible improvement and will probably cause a failure once or twice a year.

  4. #4
    Join Date
    Oct 2004
    Location
    Jämtland, Sweden
    Posts
    128

    Default Re: RAID-0 performance

    Well, I was just concerned with the performance being poor - I'm very well aware of the risks involved in having a striped array. It's just that I need all the space.
    Would JBOD work instead? I'm mainly interested in having one big drive; I chose RAID 0 because I was under the impression that it gives a performance boost, hell, it should be faster... so, is switching to JBOD possible without reformatting? Or would switching to the Sil 3114-controller make a difference, and is it at all possible without formatting the drives?

    Chieftec Bravo LBX01-B-B-SL
    580 W Hiper Type-R PSU

    Gigabyte GA-K8NXP-9 S.939
    AMD Athlon 64 Venice 3000+ at ~2470 MHz
    OCZ Gold EL PC 3700 Ram, 2.5-3-3-8 at ~456 MHz,
    2x512 MiB
    Gigabyte SilentPipe X800XL, 400/490 MHz
    Auzentech Digital X-Mystique 7.1CH DDL-RCE
    Logitech Z-5500 5.1 speakers
    1 x WD Caviar 8 MiB cache 1600JS-00MHB0 160GiB SATA-2
    3 x WD Caviar SE 16 MiB cache 2500KS-00MJBQ 250GiB SATA-2



  5. #5
    Join Date
    Nov 2003
    Location
    Minnesota, United States
    Posts
    4,543

    Default Re: RAID-0 performance

    It probably should be faster, despite the possible reasons it wouldn't be. I'd do some real-world testing, rather than trusting Sandra.

    JBOD isn't any better. You've still got all the data from one logical drive put on multiple physical drives. If one phyiscal drive fails, the logical drive is still lost. RAID 0 at least increases performance.

  6. #6
    Join Date
    Apr 2005
    Location
    Ontario, Canada
    Posts
    1,364

    Default Re: RAID-0 performance

    Raid 0 on 2 smaller drives, under 120GB, is definitely been faster for me on benchmarks and real world use. While I do not agree with Kheldar and Yawgmoth that RAID 0 is slower, I would not put 2 250GB drives in raid 0. In fact, I would recommend 80GB drives at the most and having a single drive not in the raid array big enough to keep a backup on. I have 2 120GB drives in RAID 0 and a 250GB drive I keep a complete image of them on in case of failure. I also have another 250GB drive just for important data that is also not in the raid loop. RAID 0 is when you gotta have that last inch of speed, but for reliability, their advice is spot on. Onboard motherboard controllers are really only good for raid 0 and 1. THe other options are better suited for a high performance raid controller that has some power.
    GIgabyte GA-990FXA-UD3
    AMD FX8120 @ 4GHz
    Patriot 1866MHz EL series 2X4GB DDR3
    Powercolour HD 6970 2GB w/XFX 8800GT 512MB Hybrid PhysX
    Creative X-FI titanium HD w/Technics class A 300W amp and tower speakers
    PC P&C 500W PSU
    2TB Seagate
    Coolermaster 690II w/Corsair H100 tucked under the hood

  7. #7
    Join Date
    Oct 2004
    Location
    Jämtland, Sweden
    Posts
    128

    Default Re: RAID-0 performance

    OK...
    Well... You all have some good points there... my paranoia is coming back at least :P
    I'm thinking I'll buy another 250 GiB drive, convert the array to a 0+1 array... I'll lose some space, but at least I'll be safe from failures...

    And here I thought I knew something about RAID... it's not exactly easy to find some of this info... thanks for the replys by the way!

    Chieftec Bravo LBX01-B-B-SL
    580 W Hiper Type-R PSU

    Gigabyte GA-K8NXP-9 S.939
    AMD Athlon 64 Venice 3000+ at ~2470 MHz
    OCZ Gold EL PC 3700 Ram, 2.5-3-3-8 at ~456 MHz,
    2x512 MiB
    Gigabyte SilentPipe X800XL, 400/490 MHz
    Auzentech Digital X-Mystique 7.1CH DDL-RCE
    Logitech Z-5500 5.1 speakers
    1 x WD Caviar 8 MiB cache 1600JS-00MHB0 160GiB SATA-2
    3 x WD Caviar SE 16 MiB cache 2500KS-00MJBQ 250GiB SATA-2



  8. #8
    Join Date
    Nov 2003
    Location
    Minnesota, United States
    Posts
    4,543

    Default Re: RAID-0 performance

    I never said that RAID 0 was slower. I just said it could theoretically be slower, given certain conditions, and that drives with higher spindle rates tended to be a better way to increase speed. I stand by that.

    Controllers integrated with a mainboard are more than suitable for other forms of RAID. RAID 0+1 and RAID 10 do not require much more complexity from a controller or processing power as there are no parity calculations. If the onboard SATA controller supports either of these levels, it is highly recommended. So long as you can afford 100% disk overhead.

    RAID 5 and similar forms require parity calculations, and those are best done by an integrated XOR processor (though multi-core CPUs are making this less and less true). Even RAID 5 controllers without a XOR processor are fairly expensive, so RAID with parity tends to not by an option at the consumer level.

  9. #9
    Join Date
    Apr 2005
    Location
    Ontario, Canada
    Posts
    1,364

    Default Re: RAID-0 performance

    As for knowing alot about RAID, I am still very new at it myself and I would say at best I am a beginner. Reading everything you can about it will help, and these other guys posting here definitely have more knowledge than I do. Definitely a good idea to go 0+1 though if you have data you just can't lose.
    GIgabyte GA-990FXA-UD3
    AMD FX8120 @ 4GHz
    Patriot 1866MHz EL series 2X4GB DDR3
    Powercolour HD 6970 2GB w/XFX 8800GT 512MB Hybrid PhysX
    Creative X-FI titanium HD w/Technics class A 300W amp and tower speakers
    PC P&C 500W PSU
    2TB Seagate
    Coolermaster 690II w/Corsair H100 tucked under the hood

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

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