On Tue, Jun 14, 2016 at 03:44:47PM -0400, Andrew Armenia wrote: > On Sun, Jun 12, 2016 at 10:06 PM, Anand Jain <anand.j...@oracle.com> wrote: > > > > > > On 06/03/2016 09:50 AM, Andrew Armenia wrote: > >> > >> This patch adds mount option 'chunk_width_limit=X', which when set forces > >> the chunk allocator to use only up to X devices when allocating a chunk. > >> This may help reduce the seek penalties seen in filesystems with large > >> numbers of devices. > > > > > > Have you reviewed implementations like device allocation grouping? > > Some info is in the btrfs project ideas.. > > > > https://btrfs.wiki.kernel.org/index.php/Project_ideas > > Chunk allocation groups > > Limits on number of stripes (stripe width) > > Linear chunk allocation mode > > > > (Device allocation grouping is important for enterprise storage solutions). > > > > Thanks, Anand > > I have looked at those ideas; allocation groups are what I'm ideally > after but I decided to start out small. I just spotted the dev_group > field in the device tree that appears to be currently unused, so > perhaps I will look at developing a group-aware allocator instead of > just limiting the chunk width.
I made some design notes on a generalised approach for this a while ago: http://www.spinics.net/lists/linux-btrfs/msg33782.html http://www.spinics.net/lists/linux-btrfs/msg33916.html Hugo. -- Hugo Mills | What do we want? hugo@... carfax.org.uk | Time Travel! http://carfax.org.uk/ | When do we want it? PGP: E2AB1DE4 | Irrelevant! Terminator: Genisys
signature.asc
Description: Digital signature
