patches: 1) quota initialization improvements - optimize case where quota already initialized - redesign dqptr_sem management for vfs_dq_init() 2) ext4 journalled quota make_dirty seed-up - optimize ->make_dirty() callback for journalled quota. PERFORMANCE CHART System: Laptop thinkpad t60 Core Duo T2500 (2 cores) open-close : close(open()) open-unlink : unlink(open()) write-truncate : pwrite(,,40960,0); truncate(0) - w/o quot: without quota - w quot: with journalled quota on - w sem: with dqptr_sem improvement patch - w mk_drt: with journalled make_dirty patch |----------------+----------+--------+-------+----------| | open-close | w/o quot | w quot | w sem | w mk_drt | | NR tasks 1 | 170 | 104 | 161 | 173 | | | 176 | 104 | 166 | 172 | | | 176 | 104 | 161 | 172 | | | | | | | | 4 | 80 | 42 | 81 | 79 | | | 80 | 42 | 78 | 78 | | | 80 | 41 | 80 | 79 | | | | | | | | 8 | 34 | 4 | 28 | 32 | | | 27 | 5 | 23 | 38 | | | 33 | 5 | 25 | 35 | |----------------+----------+--------+-------+----------| | open-unlink | | | | | | NR tasks 1 | 25 | 12.7 | 16 | 15.3 | | | 25 | 12.8 | 16 | 15.3 | | | 24 | 12.7 | 16 | 15.2 | | | | | | | | 4 | 8.9 | 1.1 | 3.8 | 3.8 | | | 8.9 | 1.1 | 3.3 | 3.9 | | | 8.1 | 1.1 | 3.8 | 4.0 | | | | | | | | 8 | 4.3 | 0.54 | 1.12 | 1.37 | | | 4.3 | 0.53 | 1.41 | 1.5 | | | 3.9 | 0.53 | 1.5 | 1.52 | |----------------+----------+--------+-------+----------| | write-truncate | | | | | | NR tasks 1 | 32.1 | 30.1 | 30.7 | 30.1 | | | 31.8 | 30.7 | 30.6 | 30.2 | | | 32.0 | 30.6 | 30.6 | 30.0 | | | | | | | | 4 | 8.8 | 8.45 | 8.1 | 8.3 | | | 9.0 | 9.1 | 8.2 | 9.4 | | | 9.1 | 8.9 | 9.5 | 8.4 | | | | | | | | 8 | 4.9 | 4.6 | 4.8 | 4.75 | | | 4.9 | 4.5 | 4.7 | 5.1 | | | 4.7 | 4.9 | 4.5 | 5.8 | |----------------+----------+--------+-------+----------| In fact patch-set is a first shot of larger project https://docs.google.com/fileview?id=0B2R3Idv6LrRhYjczYTA5N2MtMmUwZS00NTAzLWIwMmYtZjllODQ5OTY5YmI4&hl=en It is not clear how long does it takes to complete it. I just want to let you know what I'm think in that direction. -- 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
