| Message ID | cover.1606305501.git.fdmanana@suse.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | btrfs: some performance improvements for dbench alike workloads | expand |
On Wed, Nov 25, 2020 at 12:19:22PM +0000, fdmanana@kernel.org wrote: > From: Filipe Manana <fdmanana@suse.com> > > This patchset mostly fixes some races that result in an fsync falling back > to a transaction commit unncessarily or logging more metadata than it needs > to. It also addresses link and rename operations logging inodes that were > fsynced in the previous transaction but they don't need to be logged in > the current transaction. All these cases can be triggered often by dbench. > > Finally it avoids blocking an fsync for a significant time window when the > previous transaction is still committing. All these together resulted in > about +12% throughput with dbench and -8% maximum latency, as mentioned in > the changelog of the last patch. > > There will be more changes to improve performance with dbench or similar > workloads, but these have been cooking and testing for about one week and > are independent from what's coming next. Looking forward to more perf improvements, much appreciated. > Filipe Manana (6): > btrfs: fix race causing unnecessary inode logging during link and rename > btrfs: fix race that results in logging old extents during a fast fsync > btrfs: fix race that causes unnecessary logging of ancestor inodes > btrfs: fix race that makes inode logging fallback to transaction commit > btrfs: fix race leading to unnecessary transaction commit when logging inode > btrfs: do not block inode logging for so long during transaction commit Added to misc-next, thanks.
From: Filipe Manana <fdmanana@suse.com> This patchset mostly fixes some races that result in an fsync falling back to a transaction commit unncessarily or logging more metadata than it needs to. It also addresses link and rename operations logging inodes that were fsynced in the previous transaction but they don't need to be logged in the current transaction. All these cases can be triggered often by dbench. Finally it avoids blocking an fsync for a significant time window when the previous transaction is still committing. All these together resulted in about +12% throughput with dbench and -8% maximum latency, as mentioned in the changelog of the last patch. There will be more changes to improve performance with dbench or similar workloads, but these have been cooking and testing for about one week and are independent from what's coming next. Filipe Manana (6): btrfs: fix race causing unnecessary inode logging during link and rename btrfs: fix race that results in logging old extents during a fast fsync btrfs: fix race that causes unnecessary logging of ancestor inodes btrfs: fix race that makes inode logging fallback to transaction commit btrfs: fix race leading to unnecessary transaction commit when logging inode btrfs: do not block inode logging for so long during transaction commit fs/btrfs/ctree.h | 2 +- fs/btrfs/tree-log.c | 101 +++++++++++++++++++++++--------------------- 2 files changed, 53 insertions(+), 50 deletions(-)