fio results (many jobs, large queue depth, 64 core machine, preallocated
thin):
upstream:
READ: bw=303MiB/s (318MB/s), 15.7MiB/s-24.9MiB/s (16.5MB/s-26.1MB/s),
io=9093MiB (9535MB), run=30004-30007msec
WRITE: bw=304MiB/s (318MB/s), 15.7MiB/s-24.8MiB/s (16.5MB/s-26.0MB/s),
io=9108MiB
Thanks Mikulas,
I'll test this morning.
- Joe
On Tue, Oct 11, 2022 at 8:10 PM Mikulas Patocka wrote:
> Hi
>
> Here I'm sending updated patch 4 that fixes hang on discard. We must not
> do the optimization in dm_btree_lookup_next.
>
> Mikulas
>
>
> From: Mikulas Patocka
>
> This patch
Hi
Here I'm sending updated patch 4 that fixes hang on discard. We must not
do the optimization in dm_btree_lookup_next.
Mikulas
From: Mikulas Patocka
This patch reduces lock contention in btree walks. We modify the
functions init_ro_wpin, exit_ro_spine and ro_step so that they use