Tweaking fs.inotify.max_user_watches is not supposed to affect performance in any way. If it does, it's a bug. Noatime mounts are not supposed to affect write performance at all. It will only affect read performance of lots of small files. Ubuntu defaults to relatime (not to be confused with realtime), with an added tweak of updating it at most once a day. Setting it on root partition sounds weird, though - it may break some ancient stuff like mutt or rarely-used stuff like /tmp reapers. I believe it can be enabled for some partitions (e.g. root partition) in fstab via patched Ubiquity. It might be possible to enable noatime on udev-controlled disks using some custom rules, but I have not investigated that. The only way to make the kernel default to noatime on any mounts is a (fairly simple) kernel patch, see <a href="http://askubuntu.com/questions/59179/how-do-i-make-noatime-mounts-default">http://askubuntu.com/questions/59179/how-do-i-make-noatime-mounts-default</a> If you figure out anything for noatime by default, please let me know because elementary is also interested in using it (and as far as I can tell, we're going to use the kernel patch). Ulatencyd will get background processes out of the way (cronjobs, etc) and probably make giving realtime permissions to processes like jackd easier, though I'm not sure how it's done in US now. Maybe those two rt-permission mechanisms conflict with each other? Swappiness was already discussed at length; regarding implementation, you'll have to either find a way to tweak the OOTB /etc/sysctl.conf, or patch the kernel with a new default. -- Sergey "Shnatsel" Davidoff