1.. SPDX-License-Identifier: GPL-2.0
2
3=======================
4DAMON-based Reclamation
5=======================
6
7DAMON-based Reclamation (DAMON_RECLAIM) is a static kernel module that aimed to
8be used for proactive and lightweight reclamation under light memory pressure.
9It doesn't aim to replace the LRU-list based page_granularity reclamation, but
10to be selectively used for different level of memory pressure and requirements.
11
12Where Proactive Reclamation is Required?
13========================================
14
15On general memory over-committed systems, proactively reclaiming cold pages
16helps saving memory and reducing latency spikes that incurred by the direct
17reclaim of the process or CPU consumption of kswapd, while incurring only
18minimal performance degradation [1]_ [2]_ .
19
20Free Pages Reporting [3]_ based memory over-commit virtualization systems are
21good example of the cases.  In such systems, the guest VMs reports their free
22memory to host, and the host reallocates the reported memory to other guests.
23As a result, the memory of the systems are fully utilized.  However, the
24guests could be not so memory-frugal, mainly because some kernel subsystems and
25user-space applications are designed to use as much memory as available.  Then,
26guests could report only small amount of memory as free to host, results in
27memory utilization drop of the systems.  Running the proactive reclamation in
28guests could mitigate this problem.
29
30How It Works?
31=============
32
33DAMON_RECLAIM finds memory regions that didn't accessed for specific time
34duration and page out.  To avoid it consuming too much CPU for the paging out
35operation, a speed limit can be configured.  Under the speed limit, it pages
36out memory regions that didn't accessed longer time first.  System
37administrators can also configure under what situation this scheme should
38automatically activated and deactivated with three memory pressure watermarks.
39
40Interface: Module Parameters
41============================
42
43To use this feature, you should first ensure your system is running on a kernel
44that is built with ``CONFIG_DAMON_RECLAIM=y``.
45
46To let sysadmins enable or disable it and tune for the given system,
47DAMON_RECLAIM utilizes module parameters.  That is, you can put
48``damon_reclaim.<parameter>=<value>`` on the kernel boot command line or write
49proper values to ``/sys/modules/damon_reclaim/parameters/<parameter>`` files.
50
51Note that the parameter values except ``enabled`` are applied only when
52DAMON_RECLAIM starts.  Therefore, if you want to apply new parameter values in
53runtime and DAMON_RECLAIM is already enabled, you should disable and re-enable
54it via ``enabled`` parameter file.  Writing of the new values to proper
55parameter values should be done before the re-enablement.
56
57Below are the description of each parameter.
58
59enabled
60-------
61
62Enable or disable DAMON_RECLAIM.
63
64You can enable DAMON_RCLAIM by setting the value of this parameter as ``Y``.
65Setting it as ``N`` disables DAMON_RECLAIM.  Note that DAMON_RECLAIM could do
66no real monitoring and reclamation due to the watermarks-based activation
67condition.  Refer to below descriptions for the watermarks parameter for this.
68
69min_age
70-------
71
72Time threshold for cold memory regions identification in microseconds.
73
74If a memory region is not accessed for this or longer time, DAMON_RECLAIM
75identifies the region as cold, and reclaims it.
76
77120 seconds by default.
78
79quota_ms
80--------
81
82Limit of time for the reclamation in milliseconds.
83
84DAMON_RECLAIM tries to use only up to this time within a time window
85(quota_reset_interval_ms) for trying reclamation of cold pages.  This can be
86used for limiting CPU consumption of DAMON_RECLAIM.  If the value is zero, the
87limit is disabled.
88
8910 ms by default.
90
91quota_sz
92--------
93
94Limit of size of memory for the reclamation in bytes.
95
96DAMON_RECLAIM charges amount of memory which it tried to reclaim within a time
97window (quota_reset_interval_ms) and makes no more than this limit is tried.
98This can be used for limiting consumption of CPU and IO.  If this value is
99zero, the limit is disabled.
100
101128 MiB by default.
102
103quota_reset_interval_ms
104-----------------------
105
106The time/size quota charge reset interval in milliseconds.
107
108The charget reset interval for the quota of time (quota_ms) and size
109(quota_sz).  That is, DAMON_RECLAIM does not try reclamation for more than
110quota_ms milliseconds or quota_sz bytes within quota_reset_interval_ms
111milliseconds.
112
1131 second by default.
114
115wmarks_interval
116---------------
117
118Minimal time to wait before checking the watermarks, when DAMON_RECLAIM is
119enabled but inactive due to its watermarks rule.
120
121wmarks_high
122-----------
123
124Free memory rate (per thousand) for the high watermark.
125
126If free memory of the system in bytes per thousand bytes is higher than this,
127DAMON_RECLAIM becomes inactive, so it does nothing but only periodically checks
128the watermarks.
129
130wmarks_mid
131----------
132
133Free memory rate (per thousand) for the middle watermark.
134
135If free memory of the system in bytes per thousand bytes is between this and
136the low watermark, DAMON_RECLAIM becomes active, so starts the monitoring and
137the reclaiming.
138
139wmarks_low
140----------
141
142Free memory rate (per thousand) for the low watermark.
143
144If free memory of the system in bytes per thousand bytes is lower than this,
145DAMON_RECLAIM becomes inactive, so it does nothing but periodically checks the
146watermarks.  In the case, the system falls back to the LRU-list based page
147granularity reclamation logic.
148
149sample_interval
150---------------
151
152Sampling interval for the monitoring in microseconds.
153
154The sampling interval of DAMON for the cold memory monitoring.  Please refer to
155the DAMON documentation (:doc:`usage`) for more detail.
156
157aggr_interval
158-------------
159
160Aggregation interval for the monitoring in microseconds.
161
162The aggregation interval of DAMON for the cold memory monitoring.  Please
163refer to the DAMON documentation (:doc:`usage`) for more detail.
164
165min_nr_regions
166--------------
167
168Minimum number of monitoring regions.
169
170The minimal number of monitoring regions of DAMON for the cold memory
171monitoring.  This can be used to set lower-bound of the monitoring quality.
172But, setting this too high could result in increased monitoring overhead.
173Please refer to the DAMON documentation (:doc:`usage`) for more detail.
174
175max_nr_regions
176--------------
177
178Maximum number of monitoring regions.
179
180The maximum number of monitoring regions of DAMON for the cold memory
181monitoring.  This can be used to set upper-bound of the monitoring overhead.
182However, setting this too low could result in bad monitoring quality.  Please
183refer to the DAMON documentation (:doc:`usage`) for more detail.
184
185monitor_region_start
186--------------------
187
188Start of target memory region in physical address.
189
190The start physical address of memory region that DAMON_RECLAIM will do work
191against.  That is, DAMON_RECLAIM will find cold memory regions in this region
192and reclaims.  By default, biggest System RAM is used as the region.
193
194monitor_region_end
195------------------
196
197End of target memory region in physical address.
198
199The end physical address of memory region that DAMON_RECLAIM will do work
200against.  That is, DAMON_RECLAIM will find cold memory regions in this region
201and reclaims.  By default, biggest System RAM is used as the region.
202
203kdamond_pid
204-----------
205
206PID of the DAMON thread.
207
208If DAMON_RECLAIM is enabled, this becomes the PID of the worker thread.  Else,
209-1.
210
211Example
212=======
213
214Below runtime example commands make DAMON_RECLAIM to find memory regions that
215not accessed for 30 seconds or more and pages out.  The reclamation is limited
216to be done only up to 1 GiB per second to avoid DAMON_RECLAIM consuming too
217much CPU time for the paging out operation.  It also asks DAMON_RECLAIM to do
218nothing if the system's free memory rate is more than 50%, but start the real
219works if it becomes lower than 40%.  If DAMON_RECLAIM doesn't make progress and
220therefore the free memory rate becomes lower than 20%, it asks DAMON_RECLAIM to
221do nothing again, so that we can fall back to the LRU-list based page
222granularity reclamation. ::
223
224    # cd /sys/modules/damon_reclaim/parameters
225    # echo 30000000 > min_age
226    # echo $((1 * 1024 * 1024 * 1024)) > quota_sz
227    # echo 1000 > quota_reset_interval_ms
228    # echo 500 > wmarks_high
229    # echo 400 > wmarks_mid
230    # echo 200 > wmarks_low
231    # echo Y > enabled
232
233.. [1] https://research.google/pubs/pub48551/
234.. [2] https://lwn.net/Articles/787611/
235.. [3] https://www.kernel.org/doc/html/latest/vm/free_page_reporting.html
236