On Tue, Aug 23, 2022 at 4:51 AM Michal Hocko <mhocko@xxxxxxxx> wrote: > > On Tue 23-08-22 17:20:59, Zhaoyang Huang wrote: > > On Tue, Aug 23, 2022 at 4:33 PM Michal Hocko <mhocko@xxxxxxxx> wrote: > > > > > > On Tue 23-08-22 14:03:04, Zhaoyang Huang wrote: > > > > On Tue, Aug 23, 2022 at 1:21 PM Michal Hocko <mhocko@xxxxxxxx> wrote: > > > > > > > > > > On Tue 23-08-22 10:31:57, Zhaoyang Huang wrote: > > > [...] > > > > > > I would like to quote the comments from google side for more details > > > > > > which can also be observed from different vendors. > > > > > > "Also be advised that when you enable memcg v2 you will be using > > > > > > per-app memcg configuration which implies noticeable overhead because > > > > > > every app will have its own group. For example pagefault path will > > > > > > regress by about 15%. And obviously there will be some memory overhead > > > > > > as well. That's the reason we don't enable them in Android by > > > > > > default." > > > > > > > > > > This should be reported and investigated. Because per-application memcg > > > > > vs. memcg in general shouldn't make much of a difference from the > > > > > performance side. I can see a potential performance impact for no-memcg > > > > > vs. memcg case but even then 15% is quite a lot. > > > > Less efficiency on memory reclaim caused by multi-LRU should be one of > > > > the reason, which has been proved by comparing per-app memcg on/off. > > > > Besides, theoretically workingset could also broken as LRU is too > > > > short to compose workingset. > > > > > > Do you have any data to back these claims? Is this something that could > > > be handled on the configuration level? E.g. by applying low limit > > > protection to keep the workingset in the memory? > > I don't think so. IMO, workingset works when there are pages evicted > > from LRU and then refault which provide refault distance for pages. > > Applying memcg's protection will have all LRU out of evicted which > > make the mechanism fail. > > It is really hard to help you out without any actual data. The idea was > though to use the low limit protection to adaptively configure > respective memcgs to reduce refaults. You already have data about > refaults ready so increasing the limit for often refaulting memcgs would > reduce the trashing. Sorry for joining late. A couple years ago I tested root-memcg vs per-app memcg configurations on an Android phone. Here is a snapshot from my findings: Problem ======= We see tangible increase in major faults and workingset refaults when transitioning from root-only memory cgroup to per-application cgroups on Android. Test results ============ Results while running memory-demanding workload: root memcg per-app memcg delta workingset_refault 1771228 3874281 +118.73% workingset_nodereclaim 4543 13928 +206.58% pgpgin 13319208 20618944 +54.81% pgpgout 1739552 3080664 +77.1% pgpgoutclean 2616571 4805755 +83.67% pswpin 359211 3918716 +990.92% pswpout 1082238 5697463 +426.45% pgfree 28978393 32531010 +12.26% pgactivate 2586562 8731113 +237.56% pgdeactivate 3811074 11670051 +206.21% pgfault 38692510 46096963 +19.14% pgmajfault 441288 4100020 +829.1% pgrefill 4590451 12768165 +178.15% Results while running application cycle test (20 apps, 20 cycles): root memcg per-app memcg delta workingset_refault 10634691 11429223 +7.47% workingset_nodereclaim 37477 59033 +57.52% pgpgin 70662840 69569516 -1.55% pgpgout 2605968 2695596 +3.44% pgpgoutclean 13514955 14980610 +10.84% pswpin 1489851 3780868 +153.77% pswpout 4125547 8050819 +95.15% pgfree 99823083 105104637 +5.29% pgactivate 7685275 11647913 +51.56% pgdeactivate 14193660 21459784 +51.19% pgfault 89173166 100598528 +12.81% pgmajfault 1856172 4227190 +127.74% pgrefill 16643554 23203927 +39.42% Tests were conducted on an Android phone with 4GB RAM. Similar regression was reported a couple years ago here: https://www.spinics.net/lists/linux-mm/msg121665.html I plan on checking the difference again on newer kernels (likely 5.15) after LPC this September. > > [...] > > > A.cgroup.controllers = memory > > > A.cgroup.subtree_control = memory > > > > > > A/B.cgroup.controllers = memory > > > A/B.cgroup.subtree_control = memory > > > A/B/B1.cgroup.controllers = memory > > > > > > A/C.cgroup.controllers = memory > > > A/C.cgroup.subtree_control = "" > > > A/C/C1.cgroup.controllers = "" > > Yes for above hierarchy and configuration. > > > > > > Is your concern that C1 is charged to A/C or that you cannot actually make > > > A/C.cgroup.controllers = "" because you want to maintain memory in A? > > > Because that would be breaking the internal node constrain rule AFAICS. > > No. I just want to keep memory on B. > > That would require A to be without controllers which is not possible due > to hierarchical constrain. > > > > Or maybe you just really want a different hierarchy where > > > A == root_cgroup and want the memory acocunted in B > > > (root/B.cgroup.controllers = memory) but not in C (root/C.cgroup.controllers = "")? > > Yes. > > > > > > That would mean that C memory would be maintained on the global (root > > > memcg) LRUs which is the only internal node which is allowed to have > > > resources because it is special. > > Exactly. I would like to have all groups like C which have no parent's > > subtree_control = memory charge memory to root. Under this > > implementation, memory under enabled group will be protected by > > min/low while other groups' memory share the same LRU to have > > workingset things take effect. > > One way to achieve that would be shaping the hierarchy the following way > root > / \ > no_memcg[1] memcg[2] > |||||||| ||||| > app_cgroups app_cgroups > > with > no_memcg.subtree_control = "" > memcg.subtree_control = memory > > no? > > You haven't really described why you need per application freezer cgroup > but I suspect you want to selectively freeze applications. Is there > any obstacle to have a dedicated frozen cgroup and migrate tasks to be > frozen there? We intend for Android to gradually migrate to v2 cgroups for all controllers and given that it has to use a unified hierarchy, per-application hierarchy provides highest flexibility. That way we can control every aspect of every app without affecting others. Of course that comes with its overhead. Thanks, Suren. > -- > Michal Hocko > SUSE Labs
