I'm travelling for a conference at the moment so these patches are not tested but with the ongoing NUMA migration work I figured it was best to post these sooner rather than later. This series adds vmstat counters and tracepoints for migration, compaction and autonuma. Using them it's possible to create a basic cost model to estimate the overhead due to compaction or autonuma. Using the stats it is also possible to measure if a workload is converging on autonuma or not and potentially measure how quickly it is converging. Ideally the same stats would be available for schednuma but I did not review the series when it was last posted in July and had not seen a recent posting. I only recently heard they were in the -tip tree but will not get the chance to look at them until I've finished travelling in a weeks time. If schednuma had similar stats it would then be possible to compare schednuma and autonuma in terms of how quickly a workload converges with either approach. include/linux/migrate.h | 14 +++++++++- include/linux/vm_event_item.h | 12 ++++++++- include/trace/events/migrate.h | 52 ++++++++++++++++++++++++++++++++++++++++ mm/autonuma.c | 22 +++++++++++++---- mm/compaction.c | 15 +++++++---- mm/memory-failure.c | 3 +- mm/memory_hotplug.c | 3 +- mm/mempolicy.c | 6 +++- mm/migrate.c | 16 ++++++++++- mm/page_alloc.c | 3 +- mm/vmstat.c | 16 ++++++++++-- 11 files changed, 139 insertions(+), 23 deletions(-) create mode 100644 include/trace/events/migrate.h -- 1.7.7 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>