Hello, this is a second iteration of my series improving handling of sync syscall. Since previous submission I have incorporated Christoph's suggestions and also measured some numbers. Unpatched kernel Patched kernel time for (( i = 0; i < 100; i++ )); do sync; done 1.130400 +- 0.027369 1.073200 +- 0.040848 create 20000 4k files and sync (XFS) 155.995600 +- 1.879084 155.942200 +- 1.843881 create 20000 4k files and sync (ext3) 154.597200 +- 1.556965 153.109800 +- 1.339094 run while true; do time sync; sleep 5; done while creating 32 MB files in parallel (xfs) 6.466727 +- 0.710062 5.716000 +- 0.574406 run while true; do time sync; sleep 5; done while creating 32 MB files in parallel (ext3) 15.969909 +- 3.732143 7.662909 +- 2.268830 We see that when we are not writing in parallel to sync (the last test) there are no big improvements. When writing while running sync, ext3 got a nice boost from saving one pass over inodes. XFS got improved only slightly. Comments are welcome. Honza -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
