Increase bmds->cur_dirty after submit io, so reduce the frequency
involve into blk_drain, and improve the performance obviously
when block migration.
The performance test result of this patch:
During the block dirty save phase, this patch improve guest os IOPS
from 4.0K to 9.5K. and improve the migration speed from
505856 rsec/s to 855756 rsec/s.
Signed-off-by: Lidong Chen <jemmy858585@gmail.com>
---
migration/block.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/migration/block.c b/migration/block.c
index 6741228..7734ff7 100644
--- a/migration/block.c
+++ b/migration/block.c
@@ -576,6 +576,9 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds,
}
bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors);
+ sector += nr_sectors;
+ bmds->cur_dirty = sector;
+
break;
}
sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
--
1.8.3.1
Am 15.03.2017 um 04:37 hat Lidong Chen geschrieben: > Increase bmds->cur_dirty after submit io, so reduce the frequency > involve into blk_drain, and improve the performance obviously > when block migration. > > The performance test result of this patch: > > During the block dirty save phase, this patch improve guest os IOPS > from 4.0K to 9.5K. and improve the migration speed from > 505856 rsec/s to 855756 rsec/s. > > Signed-off-by: Lidong Chen <jemmy858585@gmail.com> > --- > migration/block.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/migration/block.c b/migration/block.c > index 6741228..7734ff7 100644 > --- a/migration/block.c > +++ b/migration/block.c > @@ -576,6 +576,9 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds, > } > > bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors); > + sector += nr_sectors; sector += BDRV_SECTORS_PER_DIRTY_CHUNK would be acceptable here... > + bmds->cur_dirty = sector; > + > break; > } > sector += BDRV_SECTORS_PER_DIRTY_CHUNK; > bmds->cur_dirty = sector; > } ...which would make it the same code as the rest of the loop. It could be nicer to reuse the same code and just set a bool when a request is submitted so that the loop terminates after that. But that's just a matter of taste, so I'll still give: Reviewed-by: Kevin Wolf <kwolf@redhat.com>
On Wed, 03/15 11:37, Lidong Chen wrote: > Increase bmds->cur_dirty after submit io, so reduce the frequency > involve into blk_drain, and improve the performance obviously > when block migration. > > The performance test result of this patch: > > During the block dirty save phase, this patch improve guest os IOPS > from 4.0K to 9.5K. and improve the migration speed from > 505856 rsec/s to 855756 rsec/s. > > Signed-off-by: Lidong Chen <jemmy858585@gmail.com> > --- > migration/block.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/migration/block.c b/migration/block.c > index 6741228..7734ff7 100644 > --- a/migration/block.c > +++ b/migration/block.c > @@ -576,6 +576,9 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds, > } > > bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors); > + sector += nr_sectors; > + bmds->cur_dirty = sector; > + > break; > } > sector += BDRV_SECTORS_PER_DIRTY_CHUNK; > -- > 1.8.3.1 > Nice catch above all, thank you! Reviewed-by: Fam Zheng <famz@redhat.com>
* Fam Zheng (famz@redhat.com) wrote: > On Wed, 03/15 11:37, Lidong Chen wrote: > > Increase bmds->cur_dirty after submit io, so reduce the frequency > > involve into blk_drain, and improve the performance obviously > > when block migration. > > > > The performance test result of this patch: > > > > During the block dirty save phase, this patch improve guest os IOPS > > from 4.0K to 9.5K. and improve the migration speed from > > 505856 rsec/s to 855756 rsec/s. > > > > Signed-off-by: Lidong Chen <jemmy858585@gmail.com> > > --- > > migration/block.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/migration/block.c b/migration/block.c > > index 6741228..7734ff7 100644 > > --- a/migration/block.c > > +++ b/migration/block.c > > @@ -576,6 +576,9 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds, > > } > > > > bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors); > > + sector += nr_sectors; > > + bmds->cur_dirty = sector; > > + > > break; > > } > > sector += BDRV_SECTORS_PER_DIRTY_CHUNK; > > -- > > 1.8.3.1 > > > > Nice catch above all, thank you! > > Reviewed-by: Fam Zheng <famz@redhat.com> Are you taking that via a block pull? Dave -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
On Wed, 03/15 17:31, Dr. David Alan Gilbert wrote: > * Fam Zheng (famz@redhat.com) wrote: > > On Wed, 03/15 11:37, Lidong Chen wrote: > > > Increase bmds->cur_dirty after submit io, so reduce the frequency > > > involve into blk_drain, and improve the performance obviously > > > when block migration. > > > > > > The performance test result of this patch: > > > > > > During the block dirty save phase, this patch improve guest os IOPS > > > from 4.0K to 9.5K. and improve the migration speed from > > > 505856 rsec/s to 855756 rsec/s. > > > > > > Signed-off-by: Lidong Chen <jemmy858585@gmail.com> > > > --- > > > migration/block.c | 3 +++ > > > 1 file changed, 3 insertions(+) > > > > > > diff --git a/migration/block.c b/migration/block.c > > > index 6741228..7734ff7 100644 > > > --- a/migration/block.c > > > +++ b/migration/block.c > > > @@ -576,6 +576,9 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds, > > > } > > > > > > bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors); > > > + sector += nr_sectors; > > > + bmds->cur_dirty = sector; > > > + > > > break; > > > } > > > sector += BDRV_SECTORS_PER_DIRTY_CHUNK; > > > -- > > > 1.8.3.1 > > > > > > > Nice catch above all, thank you! > > > > Reviewed-by: Fam Zheng <famz@redhat.com> > > Are you taking that via a block pull? I can do that, but I'm not sure whether it should go to 2.9. This is a performance improvement, which usually doesn't qualify as bug fixes. But this also looks like a mistake in original code. Fam
© 2016 - 2024 Red Hat, Inc.