Re: [ceph-users] High ops/s with kRBD and "--object-size 32M"

2016-11-29 Thread Nick Fisk
From: ceph-users [mailto:ceph-users-boun...@lists.ceph.com] On Behalf Of Alex Gorbachev Sent: 29 November 2016 04:24 To: Francois Blondel <fblon...@intelliad.de>; Ilya Dryomov <idryo...@gmail.com> Cc: ceph-us...@ceph.com Subject: Re: [ceph-users] High ops/s with kRBD and "

Re: [ceph-users] High ops/s with kRBD and "--object-size 32M"

2016-11-28 Thread Alex Gorbachev
On Mon, Nov 28, 2016 at 2:59 PM Ilya Dryomov wrote: > On Mon, Nov 28, 2016 at 6:20 PM, Francois Blondel > wrote: > > Hi *, > > > > I am currently testing different scenarios to try to optimize sequential > > read and write speeds using Kernel RBD. > >

Re: [ceph-users] High ops/s with kRBD and "--object-size 32M"

2016-11-28 Thread Ilya Dryomov
On Mon, Nov 28, 2016 at 6:20 PM, Francois Blondel wrote: > Hi *, > > I am currently testing different scenarios to try to optimize sequential > read and write speeds using Kernel RBD. > > I have two block devices created with : > rbd create block1 --size 500G --pool rbd

Re: [ceph-users] High ops/s with kRBD and "--object-size 32M"

2016-11-28 Thread Jason Dillaman
To optimize for non-direct, sequential IO, you'd actually most likely be better off with smaller RBD object sizes. The rationale is that each backing object is handled by a single PG and by using smaller objects, you can distribute the IO load to more PGs (and associated OSDs) in parallel. The 4MB

[ceph-users] High ops/s with kRBD and "--object-size 32M"

2016-11-28 Thread Francois Blondel
Hi *, I am currently testing different scenarios to try to optimize sequential read and write speeds using Kernel RBD. I have two block devices created with : rbd create block1 --size 500G --pool rbd --image-feature layering rbd create block132m --size 500G --pool rbd --image-feature