1 ===============================
2 Numa policy hit/miss statistics
3 ===============================
4
5 /sys/devices/system/node/node*/numastat
6
7 All units are pages. Hugepages have separate c
8
9 The numa_hit, numa_miss and numa_foreign count
10 are able to allocate memory from nodes they pr
11 is incremented on the preferred node, otherwis
12 the preferred node and numa_miss on the node w
13
14 Usually preferred node is the one local to the
15 but restrictions such as mempolicies can chang
16 counters based on CPU local node. local_node i
17 incremented on allocation from a node by CPU o
18 similar to numa_miss and is incremented on the
19 from a CPU from a different node. Note there i
20 numa_foreign.
21
22 In more detail:
23
24 =============== ==============================
25 numa_hit A process wanted to allocate m
26 and succeeded.
27
28 numa_miss A process wanted to allocate m
29 but ended up with memory from
30
31 numa_foreign A process wanted to allocate o
32 but ended up with memory from
33
34 local_node A process ran on this node's C
35 and got memory from this node.
36
37 other_node A process ran on a different n
38 and got memory from this node.
39
40 interleave_hit Interleaving wanted to allocat
41 and succeeded.
42 =============== ==============================
43
44 For easier reading you can use the numastat ut
45 (http://oss.sgi.com/projects/libnuma/). Note t
46 well right now on machines with a small number
47
48 Note that on systems with memoryless nodes (wh
49 memory) the numa_hit, numa_miss and numa_forei
50 heavily. In the current kernel implementation,
51 memoryless node (i.e. because it is running o
52 implementation actually treats one of the near
53 preferred node. As a result, such allocation w
54 counter on the memoryless node, and will skew
55 numa_foreign statistics of the nearest node.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.