Test results of dm-ioband
Takabe-san, thank you for testing dm-ioband, and I'm glad to hear that
it worked on your computer.
Agk, as the result shows, dm-ioband is stable and works well. I hope
it gets merged into upstream, but does dm-ioband need to be tested or
to be reviewed by someone else?
BTW, I've posted some dm-ioband patches, but the current dm quilt tree
doesn't reflect them. Is there any problem with my way of posting?
From: Akio Takebe <email@example.com>
Subject: Test results of dm-ioband
Date: Thu, 19 Mar 2009 16:51:55 +0900
> Hi, all
> Nice to meet you. I'm new to device-mapper.
> I'd like to use dm-ioband because it would be useful for virtual environment(kvm, xen).
> So I tested dm-ioband, and my test results are below.
> I hope it's a help that dm-ioband is committed to mainline.
> Any comments are welcome.
> Tested-by: Akio Takebe <firstname.lastname@example.org>
> Test Discription:
> Setup infomation:
> 1. compile and loading module
> 2. setup dmioband
> # echo "0 $(blockdev --getsize /dev/sdb7) ioband /dev/sdb7 1 0 0 none" "weight 0 :40" | dmsetup create ioband1
> # echo "0 $(blockdev --getsize /dev/sdb8) ioband /dev/sdb8 1 0 0 none" "weight 0 :10" | dmsetup create ioband2
> # mkfs.ext3 /dev/mapper/ioband1
> # mkfs.ext3 /dev/mapper/ioband2
> # mount /dev/mapper/ioband1 /mnt1
> # mount /dev/mapper/ioband2 /mnt2
> Test Items
> - load/unload module : OK
> - dmsetup targets show ioband: OK
> - Performance test results (fio)
> Summary of results
> throughput [KiB/s] % of throughput of
> device weight(%) read write total ioband1 + ioband2
> ioband1 40(80%) 328 335 663 79.59%
> ioband2 10(20%) 87 83 170 20.41%
> - I/O load (dd):
> I used some I/O load with dd command.
> I confirmed that the filesystem didn't break down.
> The files written by dd were also sane.
> Best Regards,
> Akio Takebe
dm-devel mailing list