[lug] Promise Vtrak performance
George Sexton
gsexton at mhsoftware.com
Sun Nov 25 12:51:23 MST 2007
Rob Nagler wrote:
> On 11/25/07, George Sexton wrote:
>> It would be more helpful if instead of saying performance was abysmal,
>> you gave us some numbers out of a benchmark program like bonnie++.
>
> I did say "hours" and "days" to run "du".
Sorry. Took the days part as hyperbole.
>
>> SW raid 1 runs around 50MB/Sec, while HW RAID 1 runs around 60MB/Sec.
>
> I would think these numbers are very hardware dependent.
Yup. These are numbers that I got off of a system that I own. SW RAID
numbers are using the integrated ICH7R controller, and HW RAID is the
3Ware 9550U-4LP in a 133MHz PCI slot.
So, the HW raid is maybe 1/3rd again faster (actually the numbers were
47MB versus 63MB). The CPU utilization was virtually identical.
>
>> What is yours doing?
>
> I couldn't get bonnie++ to compile on RH7.2, and I can't move the Vtrak to a
> RHEL4.5 system right now.
>
> Here's my own benchmark. /blob is the Vtrak. /usr is on a PERC3 with
> 10K SCSI RAID10. du /usr runs 42 faster than du /blob when cached,
> and uncached 10.6x faster. Even though the system time on du /usr is
> twice as much than du /blob. See results below.
Something still doesn't seem right. Are you sure the array is healthy
and not running in degraded mode?
>
> /blob is not the backup partition, obviously. On the backup partition, du
> takes days to run.
>
> Rob
> ------------------------------------------------------------------------------------------------
>
> # /usr/bin/time du /blob > /dev/null
> 0.01user 0.28system 1:25.97elapsed 0%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+58minor)pagefaults 0swaps
> # /usr/bin/time du /blob > /dev/null
> 0.01user 0.19system 0:26.73elapsed 0%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+59minor)pagefaults 0swaps
> # /usr/bin/time du /blob > /dev/null
> 0.03user 0.18system 0:25.02elapsed 0%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+58minor)pagefaults 0swaps
> # /usr/bin/time du /usr > /dev/null
> 0.08user 0.48system 0:09.10elapsed 6%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+97minor)pagefaults 0swaps
> # /usr/bin/time du /usr > /dev/null
> 0.05user 0.47system 0:00.60elapsed 86%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+96minor)pagefaults 0swaps
> # /usr/bin/time du /usr > /dev/null
> 0.14user 0.35system 0:00.58elapsed 84%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (138major+96minor)pagefaults 0swaps
> # df -i
> Filesystem Inodes IUsed IFree IUse% Mounted on
> /dev/sda3 35569664 81714 35487950 1% /
> /dev/sdd1 1430592 18601 1411991 2% /blob
> # df
> Filesystem 1k-blocks Used Available Use% Mounted on
> /dev/sda3 280046492 1515812 264305124 1% /
> /dev/sdd1 1952811400 40832356 1911979044 3% /blob
> _______________________________________________
> Web Page: http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
>
More information about the LUG
mailing list