Patch series "workingset protection/detection on the anonymous LRU list", v7.
* PROBLEM
In current implementation, newly created or swap-in anonymous page is
started on the active list. Growing the active list results in
rebalancing active/inactive list so old pages on the active list are
demoted to the inactive list. Hence, hot page on the active list isn't
protected at all.
Following is an example of this situation.
Assume that 50 hot pages on active list and system can contain total 100
pages. Numbers denote the number of pages on active/inactive list (active
| inactive). (h) stands for hot pages and (uo) stands for used-once
pages.
1. 50 hot pages on active list
50(h) | 0
2. workload: 50 newly created (used-once) pages
50(uo) | 50(h)
3. workload: another 50 newly created (used-once) pages
50(uo) | 50(uo), swap-out 50(h)
As we can see, hot pages are swapped-out and it would cause swap-in later.
* SOLUTION
Since this is what we want to avoid, this patchset implements workingset
protection. Like as the file LRU list, newly created or swap-in anonymous
page is started on the inactive list. Also, like as the file LRU list, if
enough reference happens, the page will be promoted. This simple
modification changes the above example as following.
1. 50 hot pages on active list
50(h) | 0
2. workload: 50 newly created (used-once) pages
50(h) | 50(uo)
3. workload: another 50 newly created (used-once) pages
50(h) | 50(uo), swap-out 50(uo)
hot pages remains in the active list. :)
* EXPERIMENT
I tested this scenario on my test bed and confirmed that this problem
happens on current implementation. I also checked that it is fixed by
this patchset.
* SUBJECT
workingset detection
* PROBLEM
Later part of the patchset implements the workingset detection for the
anonymous LRU list. There is a corner case that workingset protection
could cause thrashing. If we can avoid thrashing by workingset detection,
we can get the better performance.
Following is an example of thrashing due to the workingset protection.
1. 50 hot pages on active list
50(h) | 0
2. workload: 50 newly created (will be hot) pages
50(h) | 50(wh)
3. workload: another 50 newly created (used-once) pages
50(h) | 50(uo), swap-out 50(wh)
4. workload: 50 (will be hot) pages
50(h) | 50(wh), swap-in 50(wh)
5. workload: another 50 newly created (used-once) pages
50(h) | 50(uo), swap-out 50(wh)
6. repeat 4, 5
Without workingset detection, this kind of workload cannot be promoted and
thrashing happens forever.
* SOLUTION
Therefore, this patchset implements workingset detection. All the
infrastructure for workingset detecion is already implemented, so there is
not much work to do. First, extend workingset detection code to deal with
the anonymous LRU list. Then, make swap cache handles the exceptional
value for the shadow entry. Lastly, install/retrieve the shadow value
into/from the swap cache and check the refault distance.
* EXPERIMENT
I made a test program to imitates above scenario and confirmed that
problem exists. Then, I checked that this patchset fixes it.
My test setup is a virtual machine with 8 cpus and 6100MB memory. But,
the amount of the memory that the test program can use is about 280 MB.
This is because the system uses large ram-backed swap and large ramdisk to
capture the trace.
Test scenario is like as below.
1. allocate cold memory (512MB)
2. allocate hot-1 memory (96MB)
3. activate hot-1 memory (96MB)
4. allocate another hot-2 memory (96MB)
5. access cold memory (128MB)
6. access hot-2 memory (96MB)
7. repeat 5, 6
Since hot-1 memory (96MB) is on the active list, the inactive list can
contains roughly 190MB pages. hot-2 memory's re-access interval (96+128
MB) is more 190MB, so it cannot be promoted without workingset detection
and swap-in/out happens repeatedly. With this patchset, workingset
detection works and promotion happens. Therefore, swap-in/out occurs
less.
Here is the result. (average of 5 runs)
type swap-in swap-out
base 863240 989945
patch 681565 809273
As we can see, patched kernel do less swap-in/out.
* OVERALL TEST (ebizzy using modified random function)
ebizzy is the test program that main thread allocates lots of memory and
child threads access them randomly during the given times. Swap-in will
happen if allocated memory is larger than the system memory.
The random function that represents the zipf distribution is used to make
hot/cold memory. Hot/cold ratio is controlled by the parameter. If the
parameter is high, hot memory is accessed much larger than cold one. If
the parameter is low, the number of access on each memory would be
similar. I uses various parameters in order to show the effect of
patchset on various hot/cold ratio workload.
My test setup is a virtual machine with 8 cpus, 1024 MB memory and 5120 MB
ram swap.
Result format is as following.
param: 1-1024-0.1
- 1 (number of thread)
- 1024 (allocated memory size, MB)
- 0.1 (zipf distribution alpha,
0.1 works like as roughly uniform random,
1.3 works like as small portion of memory is hot and the others are cold)
pswpin: smaller is better
std: standard deviation
improvement: negative is better
* single thread
param pswpin std improvement
base 1-1024.0-0.1 14101983.40 79441.19
prot 1-1024.0-0.1 14065875.80 136413.01 ( -0.26 )
detect 1-1024.0-0.1 13910435.60 100804.82 ( -1.36 )
base 1-1024.0-0.7 7998368.80 43469.32
prot 1-1024.0-0.7 7622245.80 88318.74 ( -4.70 )
detect 1-1024.0-0.7 7618515.20 59742.07 ( -4.75 )
base 1-1024.0-1.3 1017400.80 38756.30
prot 1-1024.0-1.3 940464.60 29310.69 ( -7.56 )
detect 1-1024.0-1.3 945511.40 24579.52 ( -7.07 )
base 1-1280.0-0.1 22895541.40 50016.08
prot 1-1280.0-0.1 22860305.40 51952.37 ( -0.15 )
detect 1-1280.0-0.1 22705565.20 93380.35 ( -0.83 )
base 1-1280.0-0.7 13717645.60 46250.65
prot 1-1280.0-0.7 12935355.80 64754.43 ( -5.70 )
detect 1-1280.0-0.7 13040232.00 63304.00 ( -4.94 )
base 1-1280.0-1.3 1654251.40 4159.68
prot 1-1280.0-1.3 1522680.60 33673.50 ( -7.95 )
detect 1-1280.0-1.3 1599207.00 70327.89 ( -3.33 )
base 1-1536.0-0.1 31621775.40 31156.28
prot 1-1536.0-0.1 31540355.20 62241.36 ( -0.26 )
detect 1-1536.0-0.1 31420056.00 123831.27 ( -0.64 )
base 1-1536.0-0.7 19620760.60 60937.60
prot 1-1536.0-0.7 18337839.60 56102.58 ( -6.54 )
detect 1-1536.0-0.7 18599128.00 75289.48 ( -5.21 )
base 1-1536.0-1.3 2378142.40 20994.43
prot 1-1536.0-1.3 2166260.60 48455.46 ( -8.91 )
detect 1-1536.0-1.3 2183762.20 16883.24 ( -8.17 )
base 1-1792.0-0.1 40259714.80 90750.70
prot 1-1792.0-0.1 40053917.20 64509.47 ( -0.51 )
detect 1-1792.0-0.1 39949736.40 104989.64 ( -0.77 )
base 1-1792.0-0.7 25704884.40 69429.68
prot 1-1792.0-0.7 23937389.00 79945.60 ( -6.88 )
detect 1-1792.0-0.7 24271902.00 35044.30 ( -5.57 )
base 1-1792.0-1.3 3129497.00 32731.86
prot 1-1792.0-1.3 2796994.40 19017.26 ( -10.62 )
detect 1-1792.0-1.3 2886840.40 33938.82 ( -7.75 )
base 1-2048.0-0.1 48746924.40 50863.88
prot 1-2048.0-0.1 48631954.40 24537.30 ( -0.24 )
detect 1-2048.0-0.1 48509419.80 27085.34 ( -0.49 )
base 1-2048.0-0.7 32046424.40 78624.22
prot 1-2048.0-0.7 29764182.20 86002.26 ( -7.12 )
detect 1-2048.0-0.7 30250315.80 101282.14 ( -5.60 )
base 1-2048.0-1.3 3916723.60 24048.55
prot 1-2048.0-1.3 3490781.60 33292.61 ( -10.87 )
detect 1-2048.0-1.3 3585002.20 44942.04 ( -8.47 )
* multi thread
param pswpin std improvement
base 8-1024.0-0.1 16219822.60 329474.01
prot 8-1024.0-0.1 15959494.00 654597.45 ( -1.61 )
detect 8-1024.0-0.1 15773790.80 502275.25 ( -2.75 )
base 8-1024.0-0.7 9174107.80 537619.33
prot 8-1024.0-0.7 8571915.00 385230.08 ( -6.56 )
detect 8-1024.0-0.7 8489484.20 364683.00 ( -7.46 )
base 8-1024.0-1.3 1108495.60 83555.98
prot 8-1024.0-1.3 1038906.20 63465.20 ( -6.28 )
detect 8-1024.0-1.3 941817.80 32648.80 ( -15.04 )
base 8-1280.0-0.1 25776114.20 450480.45
prot 8-1280.0-0.1 25430847.00 465627.07 ( -1.34 )
detect 8-1280.0-0.1 25282555.00 465666.55 ( -1.91 )
base 8-1280.0-0.7 15218968.00 702007.69
prot 8-1280.0-0.7 13957947.80 492643.86 ( -8.29 )
detect 8-1280.0-0.7 14158331.20 238656.02 ( -6.97 )
base 8-1280.0-1.3 1792482.80 30512.90
prot 8-1280.0-1.3 1577686.40 34002.62 ( -11.98 )
detect 8-1280.0-1.3 1556133.00 22944.79 ( -13.19 )
base 8-1536.0-0.1 33923761.40 575455.85
prot 8-1536.0-0.1 32715766.20 300633.51 ( -3.56 )
detect 8-1536.0-0.1 33158477.40 117764.51 ( -2.26 )
base 8-1536.0-0.7 20628907.80 303851.34
prot 8-1536.0-0.7 19329511.20 341719.31 ( -6.30 )
detect 8-1536.0-0.7 20013934.00 385358.66 ( -2.98 )
base 8-1536.0-1.3 2588106.40 130769.20
prot 8-1536.0-1.3 2275222.40 89637.06 ( -12.09 )
detect 8-1536.0-1.3 2365008.40 124412.55 ( -8.62 )
base 8-1792.0-0.1 43328279.20 946469.12
prot 8-1792.0-0.1 41481980.80 525690.89 ( -4.26 )
detect 8-1792.0-0.1 41713944.60 406798.93 ( -3.73 )
base 8-1792.0-0.7 27155647.40 536253.57
prot 8-1792.0-0.7 24989406.80 502734.52 ( -7.98 )
detect 8-1792.0-0.7 25524806.40 263237.87 ( -6.01 )
base 8-1792.0-1.3 3260372.80 137907.92
prot 8-1792.0-1.3 2879187.80 63597.26 ( -11.69 )
detect 8-1792.0-1.3 2892962.20 33229.13 ( -11.27 )
base 8-2048.0-0.1 50583989.80 710121.48
prot 8-2048.0-0.1 49599984.40 228782.42 ( -1.95 )
detect 8-2048.0-0.1 50578596.00 660971.66 ( -0.01 )
base 8-2048.0-0.7 33765479.60 812659.55
prot 8-2048.0-0.7 30767021.20 462907.24 ( -8.88 )
detect 8-2048.0-0.7 32213068.80 211884.24 ( -4.60 )
base 8-2048.0-1.3 3941675.80 28436.45
prot 8-2048.0-1.3 3538742.40 76856.08 ( -10.22 )
detect 8-2048.0-1.3 3579397.80 58630.95 ( -9.19 )
As we can see, all the cases show improvement. Especially, test case with
zipf distribution 1.3 show more improvements. It means that if there is a
hot/cold tendency in anon pages, this patchset works better.
This patch (of 6):
Current implementation of LRU management for anonymous page has some
problems. Most important one is that it doesn't protect the workingset,
that is, pages on the active LRU list. Although, this problem will be
fixed in the following patchset, the preparation is required and this
patch does it.
What following patch does is to implement workingset protection. After
the following patchset, newly created or swap-in pages will start their
lifetime on the inactive list. If inactive list is too small, there is
not enough chance to be referenced and the page cannot become the
workingset.
In order to provide the newly anonymous or swap-in pages enough chance to
be referenced again, this patch makes active/inactive LRU ratio as 1:1.
This is just a temporary measure. Later patch in the series introduces
workingset detection for anonymous LRU that will be used to better decide
if pages should start on the active and inactive list. Afterwards this
patch is effectively reverted.
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Matthew Wilcox <willy@infradead.org>
Link: http://lkml.kernel.org/r/1595490560-15117-1-git-send-email-iamjoonsoo.kim@lge.com
Link: http://lkml.kernel.org/r/1595490560-15117-2-git-send-email-iamjoonsoo.kim@lge.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
ccc5dc6734