Hi, I tried experimenting with RBD striping feature: rbd image 'volume-bd873c3f-c8c7-4270-81f8-951f65fc860c': size 50 GiB in 12800 objects order 22 (4 MiB objects) snapshot_count: 0 id: 187607b9ebf28a block_name_prefix: rbd_data.187607b9ebf28a format: 2 features: layering, striping, exclusive-lock, object-map, fast-diff, deep-flatten op_features: flags: create_timestamp: Wed Oct 6 09:04:35 2021 access_timestamp: Wed Oct 6 09:56:13 2021 modify_timestamp: Wed Oct 6 09:56:12 2021 stripe unit: 2 MiB stripe count: 2 I could see that striping was properly in place and active. This supposedly should linearly increase read and write speeds, similar to a simple RAID0 setup, but I couldn't see any tangible effect. Any ideas? :-) /Z On Thu, Oct 7, 2021 at 4:15 AM Fyodor Ustinov <ufm@xxxxxx> wrote: > Hi! > > > Btw, the first interesting find: I enabled 'rbd_balance_parent_reads' on > > the clients, and single-thread reads now scale much better, I routinely > get > > similar readings from a single disk doing 4k reads with 1 thread: > > It seems to me that this function should not give any gain in "real" load. > A bunch of read operations from different virtualisations from different > virtualization nodes should already be "spread out" across all the osd. > > WBR, > Fyodor. > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |