ZIL: Reduce maximum size of WR_COPIED to 7.5K

Benchmarks show that at certain write sizes range lock/unlock take
not so much time as extra memory copy.  The exact threshold is not
obvious due to other overheads, but it is definitely lower than
~63KB used before.  Make it configurable, defaulting at 7.5KB,
that is 8KB of nearest malloc() size minus itx and lr structs.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by:	Alexander Motin <mav@FreeBSD.org>
Sponsored by:	iXsystems, Inc.
Closes #15353
This commit is contained in:
Alexander Motin 2023-10-06 13:09:27 -04:00 committed by Brian Behlendorf
parent 3755cde22a
commit 9be8ddfb3c
2 changed files with 16 additions and 6 deletions

View File

@ -2150,6 +2150,11 @@ On very fragmented pools, lowering this
.Pq typically to Sy 36 KiB
can improve performance.
.
.It Sy zil_maxcopied Ns = Ns Sy 7680 Ns B Po 7.5 KiB Pc Pq uint
This sets the maximum number of write bytes logged via WR_COPIED.
It tunes a tradeoff between additional memory copy and possibly worse log
space efficiency vs additional range lock/unlock.
.
.It Sy zil_min_commit_timeout Ns = Ns Sy 5000 Pq u64
This sets the minimum delay in nanoseconds ZIL care to delay block commit,
waiting for more records.

View File

@ -1958,26 +1958,28 @@ zil_max_log_data(zilog_t *zilog, size_t hdrsize)
/*
* Maximum amount of log space we agree to waste to reduce number of
* WR_NEED_COPY chunks to reduce zl_get_data() overhead (~12%).
* WR_NEED_COPY chunks to reduce zl_get_data() overhead (~6%).
*/
static inline uint64_t
zil_max_waste_space(zilog_t *zilog)
{
return (zil_max_log_data(zilog, sizeof (lr_write_t)) / 8);
return (zil_max_log_data(zilog, sizeof (lr_write_t)) / 16);
}
/*
* Maximum amount of write data for WR_COPIED. For correctness, consumers
* must fall back to WR_NEED_COPY if we can't fit the entire record into one
* maximum sized log block, because each WR_COPIED record must fit in a
* single log block. For space efficiency, we want to fit two records into a
* max-sized log block.
* single log block. Below that it is a tradeoff of additional memory copy
* and possibly worse log space efficiency vs additional range lock/unlock.
*/
static uint_t zil_maxcopied = 7680;
uint64_t
zil_max_copied_data(zilog_t *zilog)
{
return ((zilog->zl_max_block_size - sizeof (zil_chain_t)) / 2 -
sizeof (lr_write_t));
uint64_t max_data = zil_max_log_data(zilog, sizeof (lr_write_t));
return (MIN(max_data, zil_maxcopied));
}
/*
@ -4226,3 +4228,6 @@ ZFS_MODULE_PARAM(zfs_zil, zil_, slog_bulk, U64, ZMOD_RW,
ZFS_MODULE_PARAM(zfs_zil, zil_, maxblocksize, UINT, ZMOD_RW,
"Limit in bytes of ZIL log block size");
ZFS_MODULE_PARAM(zfs_zil, zil_, maxcopied, UINT, ZMOD_RW,
"Limit in bytes WR_COPIED size");