[LSF/MM TOPIC][LSF/MM ATTEND] blk-mq and I/O scheduling

[LSF/MM TOPIC][LSF/MM ATTEND] blk-mq and I/O scheduling

[Andreas Herrmann]

Hi,

I'd like to participate in LSF/MM and would like to present/discuss
ideas for introducing I/O scheduling support to blk-mq.

Motiviation for this is to be able use scsi-mq even on systems that
have slow (spinning) devices attached to the SCSI stack.

I think the presentation/discussion should consist of the following

(1) short overview how blk-mq currently performs with spinning devices
    (in comparison to CFQ)

(2) information about my attempt to introduce per sw queue time-slice
    to blk-mq to mitigate the performance degradation with spinning
    devices (in comparison to CFQ)

(3) hopefully I can share information about working code for another
    approach to introduce I/O scheduling for blk-mq (which I am
    currently looking into)

(4) other ideas (e.g. toggle blk-mq per host and/or why we won't look
    into it)


Thanks,

Andreas

Re: [Lsf-pc] [LSF/MM TOPIC][LSF/MM ATTEND] blk-mq and I/O scheduling

[Christoph Hellwig]

I was still hoping we'd get your slicing patches in ASAP at least.
But there are couple more topics here, so I think it would
still be useful in that case.

Re: [Lsf-pc] [LSF/MM TOPIC][LSF/MM ATTEND] blk-mq and I/O scheduling

[Hannes Reinecke]

On 02/29/2016 12:13 AM, Christoph Hellwig wrote:
> I was still hoping we'd get your slicing patches in ASAP at least.
> But there are couple more topics here, so I think it would
> still be useful in that case.
> 
Actually I have been pondering an alternative approach.
In most cases submit_bio() / submit_bh() is called in a loop, submitting
several bios in one go.
These bios typically refer to a larger piece of memory, so merging
should be trivial here.
However, both submit_bh() and submit_bio() will never convey this
information, so they will be placed on the queue as individual bios.

If we manage to link the generated bios together we can trivially
implement merging, even for the mq case.
One idea here is to add plugging around the callers, or to allow for
linked bios to be submitted.

I see if I can make some tests here.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      zSeries & Storage
hare@suse.de			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html