Re: [PATCH mm-unstable v1 1/4] mm/mglru: fix underprotected page cache

From: Yu Zhao
Date: Mon Dec 18 2023 - 22:45:15 EST

Next message: Shawn Sung (宋孝謙): "Re: [PATCH v4 1/1] drm/mediatek: Fix errors when reporting rotation capability"
Previous message: Waiman Long: "Re: [PATCH 19/50] locking/mutex: split out mutex_types.h"
In reply to: Yu Zhao: "Re: [PATCH mm-unstable v1 1/4] mm/mglru: fix underprotected page cache"
Next in thread: Kairui Song: "Re: [PATCH mm-unstable v1 1/4] mm/mglru: fix underprotected page cache"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Mon, Dec 18, 2023 at 8:21 PM Yu Zhao <yuzhao@xxxxxxxxxx> wrote:
>
> On Mon, Dec 18, 2023 at 11:05 AM Kairui Song <ryncsn@xxxxxxxxx> wrote:
> >
> > Yu Zhao <yuzhao@xxxxxxxxxx> 于2023年12月15日周五 12:56写道：
> > >
> > > On Thu, Dec 14, 2023 at 04:51:00PM -0700, Yu Zhao wrote:
> > > > On Thu, Dec 14, 2023 at 11:38 AM Kairui Song <ryncsn@xxxxxxxxx> wrote:
> > > > >
> > > > > Yu Zhao <yuzhao@xxxxxxxxxx> 于2023年12月14日周四 11:09写道：
> > > > > > On Wed, Dec 13, 2023 at 12:59:14AM -0700, Yu Zhao wrote:
> > > > > > > On Tue, Dec 12, 2023 at 8:03 PM Kairui Song <ryncsn@xxxxxxxxx> wrote:
> > > > > > > >
> > > > > > > > Kairui Song <ryncsn@xxxxxxxxx> 于2023年12月12日周二 14:52写道：
> > > > > > > > >
> > > > > > > > > Yu Zhao <yuzhao@xxxxxxxxxx> 于2023年12月12日周二 06:07写道：
> > > > > > > > > >
> > > > > > > > > > On Fri, Dec 8, 2023 at 1:24 AM Kairui Song <ryncsn@xxxxxxxxx> wrote:
> > > > > > > > > > >
> > > > > > > > > > > Yu Zhao <yuzhao@xxxxxxxxxx> 于2023年12月8日周五 14:14写道：
> > > > > > > > > > > >
> > > > > > > > > > > > Unmapped folios accessed through file descriptors can be
> > > > > > > > > > > > underprotected. Those folios are added to the oldest generation based
> > > > > > > > > > > > on:
> > > > > > > > > > > > 1. The fact that they are less costly to reclaim (no need to walk the
> > > > > > > > > > > > rmap and flush the TLB) and have less impact on performance (don't
> > > > > > > > > > > > cause major PFs and can be non-blocking if needed again).
> > > > > > > > > > > > 2. The observation that they are likely to be single-use. E.g., for
> > > > > > > > > > > > client use cases like Android, its apps parse configuration files
> > > > > > > > > > > > and store the data in heap (anon); for server use cases like MySQL,
> > > > > > > > > > > > it reads from InnoDB files and holds the cached data for tables in
> > > > > > > > > > > > buffer pools (anon).
> > > > > > > > > > > >
> > > > > > > > > > > > However, the oldest generation can be very short lived, and if so, it
> > > > > > > > > > > > doesn't provide the PID controller with enough time to respond to a
> > > > > > > > > > > > surge of refaults. (Note that the PID controller uses weighted
> > > > > > > > > > > > refaults and those from evicted generations only take a half of the
> > > > > > > > > > > > whole weight.) In other words, for a short lived generation, the
> > > > > > > > > > > > moving average smooths out the spike quickly.
> > > > > > > > > > > >
> > > > > > > > > > > > To fix the problem:
> > > > > > > > > > > > 1. For folios that are already on LRU, if they can be beyond the
> > > > > > > > > > > > tracking range of tiers, i.e., five accesses through file
> > > > > > > > > > > > descriptors, move them to the second oldest generation to give them
> > > > > > > > > > > > more time to age. (Note that tiers are used by the PID controller
> > > > > > > > > > > > to statistically determine whether folios accessed multiple times
> > > > > > > > > > > > through file descriptors are worth protecting.)
> > > > > > > > > > > > 2. When adding unmapped folios to LRU, adjust the placement of them so
> > > > > > > > > > > > that they are not too close to the tail. The effect of this is
> > > > > > > > > > > > similar to the above.
> > > > > > > > > > > >
> > > > > > > > > > > > On Android, launching 55 apps sequentially:
> > > > > > > > > > > > Before After Change
> > > > > > > > > > > > workingset_refault_anon 25641024 25598972 0%
> > > > > > > > > > > > workingset_refault_file 115016834 106178438 -8%
> > > > > > > > > > >
> > > > > > > > > > > Hi Yu,
> > > > > > > > > > >
> > > > > > > > > > > Thanks you for your amazing works on MGLRU.
> > > > > > > > > > >
> > > > > > > > > > > I believe this is the similar issue I was trying to resolve previously:
> > > > > > > > > > > https://lwn.net/Articles/945266/
> > > > > > > > > > > The idea is to use refault distance to decide if the page should be
> > > > > > > > > > > place in oldest generation or some other gen, which per my test,
> > > > > > > > > > > worked very well, and we have been using refault distance for MGLRU in
> > > > > > > > > > > multiple workloads.
> > > > > > > > > > >
> > > > > > > > > > > There are a few issues left in my previous RFC series, like anon pages
> > > > > > > > > > > in MGLRU shouldn't be considered, I wanted to collect feedback or test
> > > > > > > > > > > cases, but unfortunately it seems didn't get too much attention
> > > > > > > > > > > upstream.
> > > > > > > > > > >
> > > > > > > > > > > I think both this patch and my previous series are for solving the
> > > > > > > > > > > file pages underpertected issue, and I did a quick test using this
> > > > > > > > > > > series, for mongodb test, refault distance seems still a better
> > > > > > > > > > > solution (I'm not saying these two optimization are mutually exclusive
> > > > > > > > > > > though, just they do have some conflicts in implementation and solving
> > > > > > > > > > > similar problem):
> > > > > > > > > > >
> > > > > > > > > > > Previous result:
> > > > > > > > > > > ==================================================================
> > > > > > > > > > > Execution Results after 905 seconds
> > > > > > > > > > > ------------------------------------------------------------------
> > > > > > > > > > > Executed Time (µs) Rate
> > > > > > > > > > > STOCK_LEVEL 2542 27121571486.2 0.09 txn/s
> > > > > > > > > > > ------------------------------------------------------------------
> > > > > > > > > > > TOTAL 2542 27121571486.2 0.09 txn/s
> > > > > > > > > > >
> > > > > > > > > > > This patch:
> > > > > > > > > > > ==================================================================
> > > > > > > > > > > Execution Results after 900 seconds
> > > > > > > > > > > ------------------------------------------------------------------
> > > > > > > > > > > Executed Time (µs) Rate
> > > > > > > > > > > STOCK_LEVEL 1594 27061522574.4 0.06 txn/s
> > > > > > > > > > > ------------------------------------------------------------------
> > > > > > > > > > > TOTAL 1594 27061522574.4 0.06 txn/s
> > > > > > > > > > >
> > > > > > > > > > > Unpatched version is always around ~500.
> > > > > > > > > >
> > > > > > > > > > Thanks for the test results!
> > > > > > > > > >
> > > > > > > > > > > I think there are a few points here:
> > > > > > > > > > > - Refault distance make use of page shadow so it can better
> > > > > > > > > > > distinguish evicted pages of different access pattern (re-access
> > > > > > > > > > > distance).
> > > > > > > > > > > - Throttled refault distance can help hold part of workingset when
> > > > > > > > > > > memory is too small to hold the whole workingset.
> > > > > > > > > > >
> > > > > > > > > > > So maybe part of this patch and the bits of previous series can be
> > > > > > > > > > > combined to work better on this issue, how do you think?
> > > > > > > > > >
> > > > > > > > > > I'll try to find some time this week to look at your RFC. It'd be a
> > > > > > > >
> > > > > > > > Hi Yu,
> > > > > > > >
> > > > > > > > I'm working on V4 of the RFC now, which just update some comments, and
> > > > > > > > skip anon page re-activation in refault path for mglru which was not
> > > > > > > > very helpful, only some tiny adjustment.
> > > > > > > > And I found it easier to test with fio, using following test script:
> > > > > > > >
> > > > > > > > #!/bin/bash
> > > > > > > > swapoff -a
> > > > > > > >
> > > > > > > > modprobe brd rd_nr=1 rd_size=16777216
> > > > > > > > mkfs.ext4 /dev/ram0
> > > > > > > > mount /dev/ram0 /mnt
> > > > > > > >
> > > > > > > > mkdir -p /sys/fs/cgroup/benchmark
> > > > > > > > cd /sys/fs/cgroup/benchmark
> > > > > > > >
> > > > > > > > echo 4G > memory.max
> > > > > > > > echo $$ > cgroup.procs
> > > > > > > > echo 3 > /proc/sys/vm/drop_caches
> > > > > > > >
> > > > > > > > fio -name=mglru --numjobs=12 --directory=/mnt --size=1024m \
> > > > > > > > --buffered=1 --ioengine=io_uring --iodepth=128 \
> > > > > > > > --iodepth_batch_submit=32 --iodepth_batch_complete=32 \
> > > > > > > > --rw=randread --random_distribution=zipf:0.5 --norandommap \
> > > > > > > > --time_based --ramp_time=5m --runtime=5m --group_reporting
> > > > > > > >
> > > > > > > > zipf:0.5 is used here to simulate a cached read with slight bias
> > > > > > > > towards certain pages.
> > > > > > > > Unpatched 6.7-rc4:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=6548MiB/s (6866MB/s), 6548MiB/s-6548MiB/s
> > > > > > > > (6866MB/s-6866MB/s), io=1918GiB (2060GB), run=300001-300001msec
> > > > > > > >
> > > > > > > > Patched with RFC v4:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=7270MiB/s (7623MB/s), 7270MiB/s-7270MiB/s
> > > > > > > > (7623MB/s-7623MB/s), io=2130GiB (2287GB), run=300001-300001msec
> > > > > > > >
> > > > > > > > Patched with this series:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=7098MiB/s (7442MB/s), 7098MiB/s-7098MiB/s
> > > > > > > > (7442MB/s-7442MB/s), io=2079GiB (2233GB), run=300002-300002msec
> > > > > > > >
> > > > > > > > MGLRU off:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=6525MiB/s (6842MB/s), 6525MiB/s-6525MiB/s
> > > > > > > > (6842MB/s-6842MB/s), io=1912GiB (2052GB), run=300002-300002msec
> > > > > > > >
> > > > > > > > - If I change zipf:0.5 to random:
> > > > > > > > Unpatched 6.7-rc4:
> > > > > > > > Patched with this series:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=5975MiB/s (6265MB/s), 5975MiB/s-5975MiB/s
> > > > > > > > (6265MB/s-6265MB/s), io=1750GiB (1879GB), run=300002-300002msec
> > > > > > > >
> > > > > > > > Patched with RFC v4:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=5987MiB/s (6278MB/s), 5987MiB/s-5987MiB/s
> > > > > > > > (6278MB/s-6278MB/s), io=1754GiB (1883GB), run=300001-300001msec
> > > > > > > >
> > > > > > > > Patched with this series:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=5839MiB/s (6123MB/s), 5839MiB/s-5839MiB/s
> > > > > > > > (6123MB/s-6123MB/s), io=1711GiB (1837GB), run=300001-300001msec
> > > > > > > >
> > > > > > > > MGLRU off:
> > > > > > > > Run status group 0 (all jobs):
> > > > > > > > READ: bw=5689MiB/s (5965MB/s), 5689MiB/s-5689MiB/s
> > > > > > > > (5965MB/s-5965MB/s), io=1667GiB (1790GB), run=300003-300003msec
> > > > > > > >
> > > > > > > > fio uses ramdisk so LRU accuracy will have smaller impact. The Mongodb
> > > > > > > > test I provided before uses a SATA SSD so it will have a much higher
> > > > > > > > impact. I'll provides a script to setup the test case and run it, it's
> > > > > > > > more complex to setup than fio since involving setting up multiple
> > > > > > > > replicas and auth and hundreds of GB of test fixtures, I'm currently
> > > > > > > > occupied by some other tasks but will try best to send them out as
> > > > > > > > soon as possible.
> > > > > > >
> > > > > > > Thanks! Apparently your RFC did show better IOPS with both access
> > > > > > > patterns, which was a surprise to me because it had higher refaults
> > > > > > > and usually higher refautls result in worse performance.
> > >
> > > And thanks for providing the refaults I requested for -- your data
> > > below confirms what I mentioned above:
> > >
> > > For fio:
> > > Your RFC This series Change
> > > workingset_refault_file 628192729 596790506 -5%
> > > IOPS 1862k 1830k -2%
> > >
> > > For MongoDB:
> > > Your RFC This series Change
> > > workingset_refault_anon 10512 35277 +30%
> > > workingset_refault_file 22751782 20335355 -11%
> > > total 22762294 20370632 -11%
> > > TPS 0.09 0.06 -33%
> > >
> > > For MongoDB, this series should be a big win (but apparently it's not),
> > > especially when using zram, since an anon refault should be a lot
> > > cheaper than a file refault.
> > >
> > > So, I'm baffled...
> > >
> > > One important detail I forgot to mention: based on your data from
> > > lru_gen_full, I think there is another difference between our Kconfigs:
> > >
> > > Your Kconfig My Kconfig Max possible
> > > LRU_REFS_WIDTH 1 2 2
> >
> > Hi Yu,
> >
> > Thanks for the info, my fault, I forgot to update my config as I was
> > testing some other features.
> > Buf after I changed LRU_REFS_WIDTH to 2 by disabling IDLE_PAGE, thing
> > got much worse for MongoDB test:
> >
> > With LRU_REFS_WIDTH == 2:
> >
> > This patch:
> > ==================================================================
> > Execution Results after 919 seconds
> > ------------------------------------------------------------------
> > Executed Time (µs) Rate
> > STOCK_LEVEL 488 27598136201.9 0.02 txn/s
> > ------------------------------------------------------------------
> > TOTAL 488 27598136201.9 0.02 txn/s
> >
> > memcg 86 /system.slice/docker-1c3a90be9f0a072f5719332419550cd0e1455f2cd5863bc2780ca4d3f913ece5.scope
> > node 0
> > 1 948187 0x 0x
> > 0 0 0 0 0
> > 0 0·
> > 1 0 0 0 0
> > 0 0·
> > 2 0 0 0 0
> > 0 0·
> > 3 0 0 0 0
> > 0 0·
> > 0 0 0 0
> > 0 0·
> > 2 948187 0 6051788·
> > 0 0r 0e 0p 11916r
> > 66442e 0p
> > 1 0r 0e 0p 903r
> > 16888e 0p
> > 2 0r 0e 0p 459r
> > 9764e 0p
> > 3 0r 0e 0p 0r
> > 0e 2874p
> > 0 0 0 0
> > 0 0·
> > 3 948187 1353160 6351·
> > 0 0 0 0 0
> > 0 0·
> > 1 0 0 0 0
> > 0 0·
> > 2 0 0 0 0
> > 0 0·
> > 3 0 0 0 0
> > 0 0·
> > 0 0 0 0
> > 0 0·
> > 4 73045 23573 12·
> > 0 0R 0T 0 3498607R
> > 4868605T 0·
> > 1 0R 0T 0 3012246R
> > 3270261T 0·
> > 2 0R 0T 0 2498608R
> > 2839104T 0·
> > 3 0R 0T 0 0R
> > 1983947T 0·
> > 1486579L 0O 1380614Y 2945N
> > 2945F 2734A
> >
> > workingset_refault_anon 0
> > workingset_refault_file 18130598
> >
> > total used free shared buff/cache available
> > Mem: 31978 6705 312 20 24960 24786
> > Swap: 31977 4 31973
> >
> > RFC:
> > ==================================================================
> > Execution Results after 908 seconds
> > ------------------------------------------------------------------
> > Executed Time (µs) Rate
> > STOCK_LEVEL 2252 27159962888.2 0.08 txn/s
> > ------------------------------------------------------------------
> > TOTAL 2252 27159962888.2 0.08 txn/s
> >
> > workingset_refault_anon 22585
> > workingset_refault_file 22715256
> >
> > memcg 66 /system.slice/docker-0989446ff78106e32d3f400a0cf371c9a703281bded86d6d6bb1af706ebb25da.scope
> > node 0
> > 22 563007 2274 1198225·
> > 0 0r 1e 0p 0r
> > 697076e 0p
> > 1 0r 0e 0p 0r
> > 0e 325661p
> > 2 0r 0e 0p 0r
> > 0e 888728p
> > 3 0r 0e 0p 0r
> > 0e 3602238p
> > 0 0 0 0
> > 0 0·
> > 23 532222 7525 4948747·
> > 0 0 0 0 0
> > 0 0·
> > 1 0 0 0 0
> > 0 0·
> > 2 0 0 0 0
> > 0 0·
> > 3 0 0 0 0
> > 0 0·
> > 0 0 0 0
> > 0 0·
> > 24 500367 1214667 3292·
> > 0 0 0 0 0
> > 0 0·
> > 1 0 0 0 0
> > 0 0·
> > 2 0 0 0 0
> > 0 0·
> > 3 0 0 0 0
> > 0 0·
> > 0 0 0 0
> > 0 0·
> > 25 469692 40797 466·
> > 0 0R 271T 0 0R
> > 1162165T 0·
> > 1 0R 0T 0 774028R
> > 1205332T 0·
> > 2 0R 0T 0 0R
> > 932484T 0·
> > 3 0R 1T 0 0R
> > 4252158T 0·
> > 25178380L 156515O 23953602Y 59234N
> > 49391F 48664A
> >
> > total used free shared buff/cache available
> > Mem: 31978 6968 338 5 24671 24555
> > Swap: 31977 1533 30444
> >
> > Using same mongodb config (a 3 replica cluster using the same config):
> > {
> > "net": {
> > "bindIpAll": true,
> > "ipv6": false,
> > "maxIncomingConnections": 10000,
> > },
> > "setParameter": {
> > "disabledSecureAllocatorDomains": "*"
> > },
> > "replication": {
> > "oplogSizeMB": 10480,
> > "replSetName": "issa-tpcc_0"
> > },
> > "security": {
> > "keyFile": "/data/db/keyfile"
> > },
> > "storage": {
> > "dbPath": "/data/db/",
> > "syncPeriodSecs": 60,
> > "directoryPerDB": true,
> > "wiredTiger": {
> > "engineConfig": {
> > "cacheSizeGB": 5
> > }
> > }
> > },
> > "systemLog": {
> > "destination": "file",
> > "logAppend": true,
> > "logRotate": "rename",
> > "path": "/data/db/mongod.log",
> > "verbosity": 0
> > }
> > }
> >
> > The test environment have 32g memory and 16 core.
> >
> > Per my analyze, the access pattern for the mongodb test is that page
> > will be re-access long after it's evicted so PID controller won't
> > protect higher tier. That RFC will make use of the long existing
> > shadow to do feedback to PID/Gen so the result will be much better.
> > Still need more adjusting though, will try to do a rebase on top of
> > mm-unstable which includes your patch.
> >
> > I've no idea why the workingset_refault_* is higher in the better
> > case, this a clearly an IO bound workload, Memory and IO is busy while
> > CPU is not full...
> >
> > I've uploaded my local reproducer here:
> > https://github.com/ryncsn/emm-test-project/tree/master/mongo-cluster
> > https://github.com/ryncsn/py-tpcc
>
> Thanks for the repos -- I'm trying them right now. Which MongoDB
> version did you use? setup.sh didn't seem to install it.
>
> Also do you have a QEMU image? It'd be a lot easier for me to
> duplicate the exact environment by looking into it.

I ended up using docker.io/mongodb/mongodb-community-server:latest,
and it's not working:

# docker exec -it mongo-r1 mongosh --eval \
'"rs.initiate({
_id: "issa-tpcc_0",
members: [
{_id: 0, host: "mongo-r1"},
{_id: 1, host: "mongo-r2"},
{_id: 2, host: "mongo-r3"}
]
})"'
Emulate Docker CLI using podman. Create /etc/containers/nodocker to quiet msg.
Error: can only create exec sessions on running containers: container
state improper

Next message: Shawn Sung (宋孝謙): "Re: [PATCH v4 1/1] drm/mediatek: Fix errors when reporting rotation capability"
Previous message: Waiman Long: "Re: [PATCH 19/50] locking/mutex: split out mutex_types.h"
In reply to: Yu Zhao: "Re: [PATCH mm-unstable v1 1/4] mm/mglru: fix underprotected page cache"
Next in thread: Kairui Song: "Re: [PATCH mm-unstable v1 1/4] mm/mglru: fix underprotected page cache"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]