Re: Where is the performance bottleneck?

From: Holger Kiehl (Holger.Kiehl_at_dwd.de)
Date: 08/31/05

  • Next message: Arjan van de Ven: "Re: [ANNOUNCE] DSFS Network Forensic File System for Linux Patches" 
    Date:	Wed, 31 Aug 2005 19:00:12 +0000 (GMT)
To: Jens Axboe <axboe@suse.de>

    On Wed, 31 Aug 2005, Jens Axboe wrote:

    > On Wed, Aug 31 2005, Holger Kiehl wrote:
    >>> # ./oread /dev/sdX
    >>>
    >>> and it will read 128k chunks direct from that device. Run on the same
    >>> drives as above, reply with the vmstat info again.
    >>>
    >> Using kernel 2.6.12.5 again, here the results:
    >
    > [snip]
    >
    > Ok, reads as expected, like the buffered io but using less system time.
    > And you are still 1/3 off the target data rate, hmmm...
    >
    > With the reads, how does the aggregate bandwidth look when you add
    > 'clients'? Same as with writes, gradually decreasing per-device
    > throughput?
    >
    I performed the following tests with this command:

        dd if=/dev/sd?1 of=/dev/null bs=256k count=78125

    Single disk tests:

        /dev/sdc1 74.954715 MB/s
        /dev/sdg1 74.973417 MB/s

    Following disks in parallel:

        2 disks on same channel
        /dev/sdc1 75.034191 MB/s
        /dev/sdd1 74.984643 MB/s

        3 disks on same channel
        /dev/sdc1 75.027850 MB/s
        /dev/sdd1 74.976583 MB/s
        /dev/sde1 75.278276 MB/s

        4 disks on same channel
        /dev/sdc1 58.343166 MB/s
        /dev/sdd1 62.993059 MB/s
        /dev/sde1 66.940569 MB/s
        /dev/sdd1 70.986072 MB/s

        2 disks on different channels
        /dev/sdc1 74.954715 MB/s
        /dev/sdg1 74.973417 MB/s

        4 disks on different channels
        /dev/sdc1 74.959030 MB/s
        /dev/sdd1 74.877703 MB/s
        /dev/sdg1 75.009697 MB/s
        /dev/sdh1 75.028138 MB/s

        6 disks on different channels
        /dev/sdc1 49.640743 MB/s
        /dev/sdd1 55.935419 MB/s
        /dev/sde1 58.795241 MB/s
        /dev/sdg1 50.280864 MB/s
        /dev/sdh1 54.210705 MB/s
        /dev/sdi1 59.413176 MB/s

    So this looks different from writting, only as of four disks does the
    performance begin to drop.

    I just noticed, did you want me to do these test with the oread program?

    Thanks,
    Holger

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

  • Next message: Arjan van de Ven: "Re: [ANNOUNCE] DSFS Network Forensic File System for Linux Patches"

    Relevant Pages

    • Re: 2.6.4-mm2
      ... Jens Axboe wrote: ... How many disks are involve in the DM stack? ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: Time sliced CFQ io scheduler
      ... Jens Axboe wrote: ... > I think we need to end up with something that sets the machine profile ... > for the interesting disks. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: Linux 2.4.25-pre6
      ... Adaptec AIC7902 Ultra320 SCSI adapter ... Disks: ... SEAGATE ST373307LW X 1 ... To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ ...
      (Linux-Kernel)
    • CentOS 4.4-IBM Netvista Performace Problems, help needed.
      ... Digital IDE disks and NICs but did not re-install the OS, ... to be soft booted to recognize and boot from the disks. ... for the first few hours then the system time will begin to fall behind ... boxes into single user mode and run the date/hwclock loop and even in ...
      (comp.os.linux.hardware)
    • Re: HIGHMEM4G config for 1GB RAM on desktop?
      ... The problem is that the internal queue size of some 3ware controllers ... disks attached, you want to set queue_depth for each of them to ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)