On Wed, 2014-01-08 at 15:04 +-0000, Mel Gorman wrote: +AD4- On Tue, Dec 31, 2013 at 01:19:15PM +-0000, Chris Mason wrote: +AD4- +AD4- On Tue, 2013-12-31 at 13:45 +-0100, Jan Kara wrote: +AD4- +AD4- +AD4- On Tue 31-12-13 16:49:27, Zheng Liu wrote: +AD4- +AD4- +AD4- +AD4- Hi Chris, +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- On Mon, Dec 30, 2013 at 09:36:20PM +-0000, Chris Mason wrote: +AD4- +AD4- +AD4- +AD4- +AD4- Hi everyone, +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- I'd like to attend the LSF/MM conference this year. My current +AD4- +AD4- +AD4- +AD4- +AD4- discussion points include: +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- All things Btrfs+ACE- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- Adding cgroups for more filesystem resources, especially to limit the +AD4- +AD4- +AD4- +AD4- +AD4- speed dirty pages are created. +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- +AD4- Interesting. If I remember correctly, IO-less dirty throttling has been +AD4- +AD4- +AD4- +AD4- applied into upstream kernel, which can limit the speed that dirty pages +AD4- +AD4- +AD4- +AD4- are created. Does it has any defect? +AD4- +AD4- +AD4- It works as it should. But as Jeff points out, the throttling isn't +AD4- +AD4- +AD4- cgroup aware. So it can happen that one memcg is full of dirty pages and +AD4- +AD4- +AD4- reclaim has problems with reclaiming pages for it. I guess what Chris asks +AD4- +AD4- +AD4- for is that we watch number of dirty pages in each memcg and throttle +AD4- +AD4- +AD4- processes creating dirty pages in memcg which is close to its limit on +AD4- +AD4- +AD4- dirty pages. +AD4- +AD4- +AD4- +AD4- Right, the ioless dirty throttling is fantastic, but it's based on the +AD4- +AD4- BDI and you only get one of those per device. +AD4- +AD4- +AD4- +AD4- It's only partially related but we'll also need to keep in mind that +AD4- even with ioless dirty throttling that dirty+AF8-ratio and dirty+AF8-bytes have +AD4- been showing their age for a long time. dirty+AF8-ratio was fine when 20+ACU- +AD4- of memory was still a few seconds of IO but it has not been the case in a +AD4- long time. dirty+AF8-bytes is also not a great interface because it ignores the +AD4- speed of the underlying device. While proposals to fix it have been raised +AD4- in the past, no one (including me) has put themselves on the firing line +AD4- to replace that interface with something like dirty+AF8-time -- do not dirty +AD4- more pages than it takes N seconds to writeback. When/if someone clears +AD4- their table sufficiently to tackle that problem they are likely to collide +AD4- with any IO controller work. +AD4- I'm really hoping to combine these ideas a bit. There's a ton of overlap between the BDI throttling and the io controller. So, I'll definitely poke at it with a stick, we'll see what falls out. -chris -- 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
