Hi Jan: On Tue, Mar 8, 2011 at 2:31 PM, Jan Kara <jack@xxxxxxx> wrote: > > Hello, > > I'm posting second version of my IO-less balance_dirty_pages() patches. This > is alternative approach to Fengguang's patches - much simpler I believe (only > 300 lines added) - but obviously I does not provide so sophisticated control. > Fengguang is currently running some tests on my patches so that we can compare > the approaches. > > The basic idea (implemented in the third patch) is that processes throttled > in balance_dirty_pages() wait for enough IO to complete. The waiting is > implemented as follows: Whenever we decide to throttle a task in > balance_dirty_pages(), task adds itself to a list of tasks that are throttled > against that bdi and goes to sleep waiting to receive specified amount of page > IO completions. Once in a while (currently HZ/10, in patch 5 the interval is > autotuned based on observed IO rate), accumulated page IO completions are > distributed equally among waiting tasks. > > This waiting scheme has been chosen so that waiting time in > balance_dirty_pages() is proportional to > number_waited_pages * number_of_waiters. > In particular it does not depend on the total number of pages being waited for, > thus providing possibly a fairer results. > > Since last version I've implemented cleanups as suggested by Peter Zilstra. > The patches undergone more throughout testing. So far I've tested different > filesystems (ext2, ext3, ext4, xfs, nfs), also a combination of a local > filesystem and nfs. The load was either various number of dd threads or > fio with several threads each dirtying pages at different speed. > > Results and test scripts can be found at > http://beta.suse.com/private/jack/balance_dirty_pages-v2/ > See README file for some explanation of test framework, tests, and graphs. > Except for ext3 in data=ordered mode, where kjournald creates high > fluctuations in waiting time of throttled processes (and also high latencies), > the results look OK. Parallel dd threads are being throttled in the same way > (in a 2s window threads spend the same time waiting) and also latencies of > individual waits seem OK - except for ext3 they fit in 100 ms for local > filesystems. They are in 200-500 ms range for NFS, which isn't that nice but > to fix that we'd have to modify current ratelimiting scheme to take into > account on which bdi a page is dirtied. Then we could ratelimit slower BDIs > more often thus reducing latencies in individual waits... > > The results for different bandwidths fio load is interesting. There are 8 > threads dirtying pages at 1,2,4,..,128 MB/s rate. Due to different task > bdi dirty limits, what happens is that three most aggresive tasks get > throttled so they end up at bandwidths 24, 26, and 30 MB/s and the lighter > dirtiers run unthrottled. > > I'm planning to run some tests with multiple SATA drives to verify whether > there aren't some unexpected fluctuations. But currently I have some trouble > with the HW... > > As usual comments are welcome :). The design of IO-less foreground throttling of writeback in the context of memory cgroups is being discussed in the memcg patch threads (e.g., "[PATCH v6 0/9] memcg: per cgroup dirty page accounting"), but I've got another concern as well. And that's how restricting per-BDI writeback to a single task will affect proposed changes for tracking and accounting of buffered writes to the IO scheduler ("[RFC] [PATCH 0/6] Provide cgroup isolation for buffered writes", https://lkml.org/lkml/2011/3/8/332 ). It seems totally reasonable that reducing competition for write requests to a BDI -- by using the flusher thread to "handle" foreground writeout -- would increase throughput to that device. At Google, we experiemented with this in a hacked-up fashion several months ago (FG task would enqueue a work item and sleep for some period of time, wake up and see if it was below the dirty limit), and found that we were indeed getting better throughput. But if one of one's goals is to provide some sort of disk isolation based on cgroup parameters, than having at most one stream of write requests effectively neuters the IO scheduler. We saw that in practice, which led to abandoning our attempt at "IO-less throttling." One possible solution would be to put some of the disk isolation smarts into the writeback path, so the flusher thread could choose inodes with this as a criteria, but this seems ugly on its face, and makes my head hurt. Otherwise, I'm having trouble thinking of a way to do effective isolation in the IO scheduler without having competing threads -- for different cgroups -- making write requests for buffered data. Perhaps the best we could do would be to enable IO-less throttling in writeback as a config option? Thoughts? Thanks, Curt > > 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 > -- 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
