On 13/02/2025 05:42, Darrick J. Wong wrote: > On Wed, Feb 12, 2025 at 01:29:21PM +0000, Hans Holmberg wrote: >> On 12/02/2025 01:27, Darrick J. Wong wrote: >>> On Thu, Feb 06, 2025 at 07:44:55AM +0100, Christoph Hellwig wrote: >>>> From: Hans Holmberg <hans.holmberg@xxxxxxx> >>>> >>>> Add a file write life time data placement allocation scheme that aims to >>>> minimize fragmentation and thereby to do two things: >>>> >>>> a) separate file data to different zones when possible. >>>> b) colocate file data of similar life times when feasible. >>>> >>>> To get best results, average file sizes should align with the zone >>>> capacity that is reported through the XFS_IOC_FSGEOMETRY ioctl. >>>> >>>> For RocksDB using leveled compaction, the lifetime hints can improve >>>> throughput for overwrite workloads at 80% file system utilization by >>>> ~10%. >>> >>> The code changes look mostly ok, but how does it do at 40% utilization? >>> 99%? Does it reduce the amount of relocation work that the gc must do? >> >> The improvement in data placement efficiency will always be there, >> reducing the number of blocks requiring relocation by GC, but the impact >> on performance varies depending on how full the file system is. >> >> At 40% utilization there is almost no garbage collection going on, so the >> impact on throughput is not significant. At 99% the effects of better >> data placement should be higher. > > <nod> Would you mind pasting that into the commit message? > That sounds like good idea. Christoph: could you fold in the above lines into the commit message for the next iteration of the series? (Or do you want an updated patch?)
