1.. SPDX-License-Identifier: GPL-2.0
2
3======
4Design
5======
6
7Configurable Layers
8===================
9
10DAMON provides data access monitoring functionality while making the accuracy
11and the overhead controllable.  The fundamental access monitorings require
12primitives that dependent on and optimized for the target address space.  On
13the other hand, the accuracy and overhead tradeoff mechanism, which is the core
14of DAMON, is in the pure logic space.  DAMON separates the two parts in
15different layers and defines its interface to allow various low level
16primitives implementations configurable with the core logic.  We call the low
17level primitives implementations monitoring operations.
18
19Due to this separated design and the configurable interface, users can extend
20DAMON for any address space by configuring the core logics with appropriate
21monitoring operations.  If appropriate one is not provided, users can implement
22the operations on their own.
23
24For example, physical memory, virtual memory, swap space, those for specific
25processes, NUMA nodes, files, and backing memory devices would be supportable.
26Also, if some architectures or devices support special optimized access check
27primitives, those will be easily configurable.
28
29
30Reference Implementations of Address Space Specific Monitoring Operations
31=========================================================================
32
33The monitoring operations are defined in two parts:
34
351. Identification of the monitoring target address range for the address space.
362. Access check of specific address range in the target space.
37
38DAMON currently provides the implementations of the operations for the physical
39and virtual address spaces. Below two subsections describe how those work.
40
41
42VMA-based Target Address Range Construction
43-------------------------------------------
44
45This is only for the virtual address space monitoring operations
46implementation.  That for the physical address space simply asks users to
47manually set the monitoring target address ranges.
48
49Only small parts in the super-huge virtual address space of the processes are
50mapped to the physical memory and accessed.  Thus, tracking the unmapped
51address regions is just wasteful.  However, because DAMON can deal with some
52level of noise using the adaptive regions adjustment mechanism, tracking every
53mapping is not strictly required but could even incur a high overhead in some
54cases.  That said, too huge unmapped areas inside the monitoring target should
55be removed to not take the time for the adaptive mechanism.
56
57For the reason, this implementation converts the complex mappings to three
58distinct regions that cover every mapped area of the address space.  The two
59gaps between the three regions are the two biggest unmapped areas in the given
60address space.  The two biggest unmapped areas would be the gap between the
61heap and the uppermost mmap()-ed region, and the gap between the lowermost
62mmap()-ed region and the stack in most of the cases.  Because these gaps are
63exceptionally huge in usual address spaces, excluding these will be sufficient
64to make a reasonable trade-off.  Below shows this in detail::
65
66    <heap>
67    <BIG UNMAPPED REGION 1>
68    <uppermost mmap()-ed region>
69    (small mmap()-ed regions and munmap()-ed regions)
70    <lowermost mmap()-ed region>
71    <BIG UNMAPPED REGION 2>
72    <stack>
73
74
75PTE Accessed-bit Based Access Check
76-----------------------------------
77
78Both of the implementations for physical and virtual address spaces use PTE
79Accessed-bit for basic access checks.  Only one difference is the way of
80finding the relevant PTE Accessed bit(s) from the address.  While the
81implementation for the virtual address walks the page table for the target task
82of the address, the implementation for the physical address walks every page
83table having a mapping to the address.  In this way, the implementations find
84and clear the bit(s) for next sampling target address and checks whether the
85bit(s) set again after one sampling period.  This could disturb other kernel
86subsystems using the Accessed bits, namely Idle page tracking and the reclaim
87logic.  DAMON does nothing to avoid disturbing Idle page tracking, so handling
88the interference is the responsibility of sysadmins.  However, it solves the
89conflict with the reclaim logic using ``PG_idle`` and ``PG_young`` page flags,
90as Idle page tracking does.
91
92
93Address Space Independent Core Mechanisms
94=========================================
95
96Below four sections describe each of the DAMON core mechanisms and the five
97monitoring attributes, ``sampling interval``, ``aggregation interval``,
98``update interval``, ``minimum number of regions``, and ``maximum number of
99regions``.
100
101
102Access Frequency Monitoring
103---------------------------
104
105The output of DAMON says what pages are how frequently accessed for a given
106duration.  The resolution of the access frequency is controlled by setting
107``sampling interval`` and ``aggregation interval``.  In detail, DAMON checks
108access to each page per ``sampling interval`` and aggregates the results.  In
109other words, counts the number of the accesses to each page.  After each
110``aggregation interval`` passes, DAMON calls callback functions that previously
111registered by users so that users can read the aggregated results and then
112clears the results.  This can be described in below simple pseudo-code::
113
114    while monitoring_on:
115        for page in monitoring_target:
116            if accessed(page):
117                nr_accesses[page] += 1
118        if time() % aggregation_interval == 0:
119            for callback in user_registered_callbacks:
120                callback(monitoring_target, nr_accesses)
121            for page in monitoring_target:
122                nr_accesses[page] = 0
123        sleep(sampling interval)
124
125The monitoring overhead of this mechanism will arbitrarily increase as the
126size of the target workload grows.
127
128
129Region Based Sampling
130---------------------
131
132To avoid the unbounded increase of the overhead, DAMON groups adjacent pages
133that assumed to have the same access frequencies into a region.  As long as the
134assumption (pages in a region have the same access frequencies) is kept, only
135one page in the region is required to be checked.  Thus, for each ``sampling
136interval``, DAMON randomly picks one page in each region, waits for one
137``sampling interval``, checks whether the page is accessed meanwhile, and
138increases the access frequency of the region if so.  Therefore, the monitoring
139overhead is controllable by setting the number of regions.  DAMON allows users
140to set the minimum and the maximum number of regions for the trade-off.
141
142This scheme, however, cannot preserve the quality of the output if the
143assumption is not guaranteed.
144
145
146Adaptive Regions Adjustment
147---------------------------
148
149Even somehow the initial monitoring target regions are well constructed to
150fulfill the assumption (pages in same region have similar access frequencies),
151the data access pattern can be dynamically changed.  This will result in low
152monitoring quality.  To keep the assumption as much as possible, DAMON
153adaptively merges and splits each region based on their access frequency.
154
155For each ``aggregation interval``, it compares the access frequencies of
156adjacent regions and merges those if the frequency difference is small.  Then,
157after it reports and clears the aggregated access frequency of each region, it
158splits each region into two or three regions if the total number of regions
159will not exceed the user-specified maximum number of regions after the split.
160
161In this way, DAMON provides its best-effort quality and minimal overhead while
162keeping the bounds users set for their trade-off.
163
164
165Dynamic Target Space Updates Handling
166-------------------------------------
167
168The monitoring target address range could dynamically changed.  For example,
169virtual memory could be dynamically mapped and unmapped.  Physical memory could
170be hot-plugged.
171
172As the changes could be quite frequent in some cases, DAMON allows the
173monitoring operations to check dynamic changes including memory mapping changes
174and applies it to monitoring operations-related data structures such as the
175abstracted monitoring target memory area only for each of a user-specified time
176interval (``update interval``).
177