0001 .. SPDX-License-Identifier: GPL-2.0
0002
0003 =============================
0004 DAMON-based LRU-lists Sorting
0005 =============================
0006
0007 DAMON-based LRU-lists Sorting (DAMON_LRU_SORT) is a static kernel module that
0008 aimed to be used for proactive and lightweight data access pattern based
0009 (de)prioritization of pages on their LRU-lists for making LRU-lists a more
0010 trusworthy data access pattern source.
0011
0012 Where Proactive LRU-lists Sorting is Required?
0013 ==============================================
0014
0015 As page-granularity access checking overhead could be significant on huge
0016 systems, LRU lists are normally not proactively sorted but partially and
0017 reactively sorted for special events including specific user requests, system
0018 calls and memory pressure. As a result, LRU lists are sometimes not so
0019 perfectly prepared to be used as a trustworthy access pattern source for some
0020 situations including reclamation target pages selection under sudden memory
0021 pressure.
0022
0023 Because DAMON can identify access patterns of best-effort accuracy while
0024 inducing only user-specified range of overhead, proactively running
0025 DAMON_LRU_SORT could be helpful for making LRU lists more trustworthy access
0026 pattern source with low and controlled overhead.
0027
0028 How It Works?
0029 =============
0030
0031 DAMON_LRU_SORT finds hot pages (pages of memory regions that showing access
0032 rates that higher than a user-specified threshold) and cold pages (pages of
0033 memory regions that showing no access for a time that longer than a
0034 user-specified threshold) using DAMON, and prioritizes hot pages while
0035 deprioritizing cold pages on their LRU-lists. To avoid it consuming too much
0036 CPU for the prioritizations, a CPU time usage limit can be configured. Under
0037 the limit, it prioritizes and deprioritizes more hot and cold pages first,
0038 respectively. System administrators can also configure under what situation
0039 this scheme should automatically activated and deactivated with three memory
0040 pressure watermarks.
0041
0042 Its default parameters for hotness/coldness thresholds and CPU quota limit are
0043 conservatively chosen. That is, the module under its default parameters could
0044 be widely used without harm for common situations while providing a level of
0045 benefits for systems having clear hot/cold access patterns under memory
0046 pressure while consuming only a limited small portion of CPU time.
0047
0048 Interface: Module Parameters
0049 ============================
0050
0051 To use this feature, you should first ensure your system is running on a kernel
0052 that is built with ``CONFIG_DAMON_LRU_SORT=y``.
0053
0054 To let sysadmins enable or disable it and tune for the given system,
0055 DAMON_LRU_SORT utilizes module parameters. That is, you can put
0056 ``damon_lru_sort.<parameter>=<value>`` on the kernel boot command line or write
0057 proper values to ``/sys/modules/damon_lru_sort/parameters/<parameter>`` files.
0058
0059 Below are the description of each parameter.
0060
0061 enabled
0062 -------
0063
0064 Enable or disable DAMON_LRU_SORT.
0065
0066 You can enable DAMON_LRU_SORT by setting the value of this parameter as ``Y``.
0067 Setting it as ``N`` disables DAMON_LRU_SORT. Note that DAMON_LRU_SORT could do
0068 no real monitoring and LRU-lists sorting due to the watermarks-based activation
0069 condition. Refer to below descriptions for the watermarks parameter for this.
0070
0071 commit_inputs
0072 -------------
0073
0074 Make DAMON_LRU_SORT reads the input parameters again, except ``enabled``.
0075
0076 Input parameters that updated while DAMON_LRU_SORT is running are not applied
0077 by default. Once this parameter is set as ``Y``, DAMON_LRU_SORT reads values
0078 of parametrs except ``enabled`` again. Once the re-reading is done, this
0079 parameter is set as ``N``. If invalid parameters are found while the
0080 re-reading, DAMON_LRU_SORT will be disabled.
0081
0082 hot_thres_access_freq
0083 ---------------------
0084
0085 Access frequency threshold for hot memory regions identification in permil.
0086
0087 If a memory region is accessed in frequency of this or higher, DAMON_LRU_SORT
0088 identifies the region as hot, and mark it as accessed on the LRU list, so that
0089 it could not be reclaimed under memory pressure. 50% by default.
0090
0091 cold_min_age
0092 ------------
0093
0094 Time threshold for cold memory regions identification in microseconds.
0095
0096 If a memory region is not accessed for this or longer time, DAMON_LRU_SORT
0097 identifies the region as cold, and mark it as unaccessed on the LRU list, so
0098 that it could be reclaimed first under memory pressure. 120 seconds by
0099 default.
0100
0101 quota_ms
0102 --------
0103
0104 Limit of time for trying the LRU lists sorting in milliseconds.
0105
0106 DAMON_LRU_SORT tries to use only up to this time within a time window
0107 (quota_reset_interval_ms) for trying LRU lists sorting. This can be used
0108 for limiting CPU consumption of DAMON_LRU_SORT. If the value is zero, the
0109 limit is disabled.
0110
0111 10 ms by default.
0112
0113 quota_reset_interval_ms
0114 -----------------------
0115
0116 The time quota charge reset interval in milliseconds.
0117
0118 The charge reset interval for the quota of time (quota_ms). That is,
0119 DAMON_LRU_SORT does not try LRU-lists sorting for more than quota_ms
0120 milliseconds or quota_sz bytes within quota_reset_interval_ms milliseconds.
0121
0122 1 second by default.
0123
0124 wmarks_interval
0125 ---------------
0126
0127 The watermarks check time interval in microseconds.
0128
0129 Minimal time to wait before checking the watermarks, when DAMON_LRU_SORT is
0130 enabled but inactive due to its watermarks rule. 5 seconds by default.
0131
0132 wmarks_high
0133 -----------
0134
0135 Free memory rate (per thousand) for the high watermark.
0136
0137 If free memory of the system in bytes per thousand bytes is higher than this,
0138 DAMON_LRU_SORT becomes inactive, so it does nothing but periodically checks the
0139 watermarks. 200 (20%) by default.
0140
0141 wmarks_mid
0142 ----------
0143
0144 Free memory rate (per thousand) for the middle watermark.
0145
0146 If free memory of the system in bytes per thousand bytes is between this and
0147 the low watermark, DAMON_LRU_SORT becomes active, so starts the monitoring and
0148 the LRU-lists sorting. 150 (15%) by default.
0149
0150 wmarks_low
0151 ----------
0152
0153 Free memory rate (per thousand) for the low watermark.
0154
0155 If free memory of the system in bytes per thousand bytes is lower than this,
0156 DAMON_LRU_SORT becomes inactive, so it does nothing but periodically checks the
0157 watermarks. 50 (5%) by default.
0158
0159 sample_interval
0160 ---------------
0161
0162 Sampling interval for the monitoring in microseconds.
0163
0164 The sampling interval of DAMON for the cold memory monitoring. Please refer to
0165 the DAMON documentation (:doc:`usage`) for more detail. 5ms by default.
0166
0167 aggr_interval
0168 -------------
0169
0170 Aggregation interval for the monitoring in microseconds.
0171
0172 The aggregation interval of DAMON for the cold memory monitoring. Please
0173 refer to the DAMON documentation (:doc:`usage`) for more detail. 100ms by
0174 default.
0175
0176 min_nr_regions
0177 --------------
0178
0179 Minimum number of monitoring regions.
0180
0181 The minimal number of monitoring regions of DAMON for the cold memory
0182 monitoring. This can be used to set lower-bound of the monitoring quality.
0183 But, setting this too high could result in increased monitoring overhead.
0184 Please refer to the DAMON documentation (:doc:`usage`) for more detail. 10 by
0185 default.
0186
0187 max_nr_regions
0188 --------------
0189
0190 Maximum number of monitoring regions.
0191
0192 The maximum number of monitoring regions of DAMON for the cold memory
0193 monitoring. This can be used to set upper-bound of the monitoring overhead.
0194 However, setting this too low could result in bad monitoring quality. Please
0195 refer to the DAMON documentation (:doc:`usage`) for more detail. 1000 by
0196 defaults.
0197
0198 monitor_region_start
0199 --------------------
0200
0201 Start of target memory region in physical address.
0202
0203 The start physical address of memory region that DAMON_LRU_SORT will do work
0204 against. By default, biggest System RAM is used as the region.
0205
0206 monitor_region_end
0207 ------------------
0208
0209 End of target memory region in physical address.
0210
0211 The end physical address of memory region that DAMON_LRU_SORT will do work
0212 against. By default, biggest System RAM is used as the region.
0213
0214 kdamond_pid
0215 -----------
0216
0217 PID of the DAMON thread.
0218
0219 If DAMON_LRU_SORT is enabled, this becomes the PID of the worker thread. Else,
0220 -1.
0221
0222 nr_lru_sort_tried_hot_regions
0223 -----------------------------
0224
0225 Number of hot memory regions that tried to be LRU-sorted.
0226
0227 bytes_lru_sort_tried_hot_regions
0228 --------------------------------
0229
0230 Total bytes of hot memory regions that tried to be LRU-sorted.
0231
0232 nr_lru_sorted_hot_regions
0233 -------------------------
0234
0235 Number of hot memory regions that successfully be LRU-sorted.
0236
0237 bytes_lru_sorted_hot_regions
0238 ----------------------------
0239
0240 Total bytes of hot memory regions that successfully be LRU-sorted.
0241
0242 nr_hot_quota_exceeds
0243 --------------------
0244
0245 Number of times that the time quota limit for hot regions have exceeded.
0246
0247 nr_lru_sort_tried_cold_regions
0248 ------------------------------
0249
0250 Number of cold memory regions that tried to be LRU-sorted.
0251
0252 bytes_lru_sort_tried_cold_regions
0253 ---------------------------------
0254
0255 Total bytes of cold memory regions that tried to be LRU-sorted.
0256
0257 nr_lru_sorted_cold_regions
0258 --------------------------
0259
0260 Number of cold memory regions that successfully be LRU-sorted.
0261
0262 bytes_lru_sorted_cold_regions
0263 -----------------------------
0264
0265 Total bytes of cold memory regions that successfully be LRU-sorted.
0266
0267 nr_cold_quota_exceeds
0268 ---------------------
0269
0270 Number of times that the time quota limit for cold regions have exceeded.
0271
0272 Example
0273 =======
0274
0275 Below runtime example commands make DAMON_LRU_SORT to find memory regions
0276 having >=50% access frequency and LRU-prioritize while LRU-deprioritizing
0277 memory regions that not accessed for 120 seconds. The prioritization and
0278 deprioritization is limited to be done using only up to 1% CPU time to avoid
0279 DAMON_LRU_SORT consuming too much CPU time for the (de)prioritization. It also
0280 asks DAMON_LRU_SORT to do nothing if the system's free memory rate is more than
0281 50%, but start the real works if it becomes lower than 40%. If DAMON_RECLAIM
0282 doesn't make progress and therefore the free memory rate becomes lower than
0283 20%, it asks DAMON_LRU_SORT to do nothing again, so that we can fall back to
0284 the LRU-list based page granularity reclamation. ::
0285
0286 # cd /sys/modules/damon_lru_sort/parameters
0287 # echo 500 > hot_thres_access_freq
0288 # echo 120000000 > cold_min_age
0289 # echo 10 > quota_ms
0290 # echo 1000 > quota_reset_interval_ms
0291 # echo 500 > wmarks_high
0292 # echo 400 > wmarks_mid
0293 # echo 200 > wmarks_low
0294 # echo Y > enabled