More speculative prefetcher improvements

- Make prefetch distance adaptive: up to 4MB prefetch doubles for
every, hit same as before, but after that it grows by 1/8 every time
the prefetch read does not complete in time to satisfy the demand.
My tests show that 4MB is sufficient for wide NVMe pool to saturate
single reader thread at 2.5GB/s, while new 64MB maximum allows the
same thread to reach 1.5GB/s on wide HDD pool.  Further distance
increase may increase speed even more, but less dramatic and with
higher latency.

 - Allow early reuse of inactive prefetch streams: streams that never
saw hits can be reused immediately if there is a demand, while others
can be reused after 1s of inactivity, starting with the oldest.  After
2s of inactivity streams are deleted to free resources same as before.
This allows by several times increase strided read performance on HDD
pool in presence of simultaneous random reads, previously filling the
zfetch_max_streams limit for seconds and so blocking most of prefetch.

 - Always issue intermediate indirect block reads with SYNC priority.
Each of those reads if delayed for longer may delay up to 1024 other
block prefetches, that may be not good for wide pools.

Reviewed-by: Allan Jude <allan@klarasystems.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Sponsored-By: iXsystems, Inc.
Closes #13452
This commit is contained in:
Alexander Motin
2022-05-25 13:12:52 -04:00
committed by GitHub
parent 1d89b989c1
commit 6aa8c21a2a
5 changed files with 133 additions and 101 deletions
+14 -3
View File
@@ -487,7 +487,15 @@ However, this is limited by
.It Sy zfetch_array_rd_sz Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
If prefetching is enabled, disable prefetching for reads larger than this size.
.
.It Sy zfetch_max_distance Ns = Ns Sy 8388608 Ns B Po 8 MiB Pc Pq uint
.It Sy zfetch_min_distance Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
Min bytes to prefetch per stream.
Prefetch distance starts from the demand access size and quickly grows to
this value, doubling on each hit.
After that it may grow further by 1/8 per hit, but only if some prefetch
since last time haven't completed in time to satisfy demand request, i.e.
prefetch depth didn't cover the read latency or the pool got saturated.
.
.It Sy zfetch_max_distance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
Max bytes to prefetch per stream.
.
.It Sy zfetch_max_idistance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
@@ -496,8 +504,11 @@ Max bytes to prefetch indirects for per stream.
.It Sy zfetch_max_streams Ns = Ns Sy 8 Pq uint
Max number of streams per zfetch (prefetch streams per file).
.
.It Sy zfetch_min_sec_reap Ns = Ns Sy 2 Pq uint
Min time before an active prefetch stream can be reclaimed
.It Sy zfetch_min_sec_reap Ns = Ns Sy 1 Pq uint
Min time before inactive prefetch stream can be reclaimed
.
.It Sy zfetch_max_sec_reap Ns = Ns Sy 2 Pq uint
Max time before inactive prefetch stream can be deleted
.
.It Sy zfs_abd_scatter_enabled Ns = Ns Sy 1 Ns | Ns 0 Pq int
Enables ARC from using scatter/gather lists and forces all allocations to be