| Message ID | cover.1611742865.git.fdmanana@suse.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | btrfs: more performance improvements for dbench workloads | expand |
On 1/27/21 5:34 AM, fdmanana@kernel.org wrote: > From: Filipe Manana <fdmanana@suse.com> > > The following patchset brings one more batch of performance improvements > with dbench workloads, or anything that mixes file creation, file writes, > renames, unlinks, etc with fsync like dbench does. This patchset is mostly > based on avoiding logging directory inodes multiple times when not necessary, > falling back to transaction commits less frequently and often waiting less > time for transaction commits to complete. Performance results are listed in > the change log of the last patch, but in short, I've experienced a reduction > of maximum latency up to about -40% and throuhput gains up to about +6%. > > Filipe Manana (7): > btrfs: remove unnecessary directory inode item update when deleting dir entry > btrfs: stop setting nbytes when filling inode item for logging > btrfs: avoid logging new ancestor inodes when logging new inode > btrfs: skip logging directories already logged when logging all parents > btrfs: skip logging inodes already logged when logging new entries > btrfs: remove unnecessary check_parent_dirs_for_sync() > btrfs: make concurrent fsyncs wait less when waiting for a transaction commit > > fs/btrfs/file.c | 1 + > fs/btrfs/transaction.c | 39 +++++++-- > fs/btrfs/transaction.h | 2 + > fs/btrfs/tree-log.c | 195 ++++++++++++----------------------------- > 4 files changed, 92 insertions(+), 145 deletions(-) > You can add Reviewed-by: Josef Bacik <josef@toxicpanda.com> to the whole series, thanks, Josef
On Wed, Jan 27, 2021 at 10:34:53AM +0000, fdmanana@kernel.org wrote: > From: Filipe Manana <fdmanana@suse.com> > > The following patchset brings one more batch of performance improvements > with dbench workloads, or anything that mixes file creation, file writes, > renames, unlinks, etc with fsync like dbench does. This patchset is mostly > based on avoiding logging directory inodes multiple times when not necessary, > falling back to transaction commits less frequently and often waiting less > time for transaction commits to complete. Performance results are listed in > the change log of the last patch, but in short, I've experienced a reduction > of maximum latency up to about -40% and throuhput gains up to about +6%. Nice, added to misc-next, thanks.
From: Filipe Manana <fdmanana@suse.com> The following patchset brings one more batch of performance improvements with dbench workloads, or anything that mixes file creation, file writes, renames, unlinks, etc with fsync like dbench does. This patchset is mostly based on avoiding logging directory inodes multiple times when not necessary, falling back to transaction commits less frequently and often waiting less time for transaction commits to complete. Performance results are listed in the change log of the last patch, but in short, I've experienced a reduction of maximum latency up to about -40% and throuhput gains up to about +6%. Filipe Manana (7): btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit fs/btrfs/file.c | 1 + fs/btrfs/transaction.c | 39 +++++++-- fs/btrfs/transaction.h | 2 + fs/btrfs/tree-log.c | 195 ++++++++++++----------------------------- 4 files changed, 92 insertions(+), 145 deletions(-)