Commit | Line | Data |
---|---|---|
e4a5c33e | 1 | =============================== |
1da177e4 | 2 | Numa policy hit/miss statistics |
e4a5c33e | 3 | =============================== |
1da177e4 LT |
4 | |
5 | /sys/devices/system/node/node*/numastat | |
6 | ||
7 | All units are pages. Hugepages have separate counters. | |
8 | ||
77691ee9 VB |
9 | The numa_hit, numa_miss and numa_foreign counters reflect how well processes |
10 | are able to allocate memory from nodes they prefer. If they succeed, numa_hit | |
11 | is incremented on the preferred node, otherwise numa_foreign is incremented on | |
12 | the preferred node and numa_miss on the node where allocation succeeded. | |
13 | ||
14 | Usually preferred node is the one local to the CPU where the process executes, | |
15 | but restrictions such as mempolicies can change that, so there are also two | |
16 | counters based on CPU local node. local_node is similar to numa_hit and is | |
17 | incremented on allocation from a node by CPU on the same node. other_node is | |
18 | similar to numa_miss and is incremented on the node where allocation succeeds | |
19 | from a CPU from a different node. Note there is no counter analogical to | |
20 | numa_foreign. | |
21 | ||
22 | In more detail: | |
23 | ||
e4a5c33e | 24 | =============== ============================================================ |
36519c9e GL |
25 | numa_hit A process wanted to allocate memory from this node, |
26 | and succeeded. | |
27 | ||
28 | numa_miss A process wanted to allocate memory from another node, | |
29 | but ended up with memory from this node. | |
30 | ||
31 | numa_foreign A process wanted to allocate on this node, | |
77691ee9 | 32 | but ended up with memory from another node. |
36519c9e | 33 | |
77691ee9 VB |
34 | local_node A process ran on this node's CPU, |
35 | and got memory from this node. | |
36519c9e | 36 | |
77691ee9 VB |
37 | other_node A process ran on a different node's CPU |
38 | and got memory from this node. | |
36519c9e GL |
39 | |
40 | interleave_hit Interleaving wanted to allocate from this node | |
41 | and succeeded. | |
e4a5c33e | 42 | =============== ============================================================ |
1da177e4 LT |
43 | |
44 | For easier reading you can use the numastat utility from the numactl package | |
36519c9e | 45 | (http://oss.sgi.com/projects/libnuma/). Note that it only works |
1da177e4 LT |
46 | well right now on machines with a small number of CPUs. |
47 | ||
77691ee9 VB |
48 | Note that on systems with memoryless nodes (where a node has CPUs but no |
49 | memory) the numa_hit, numa_miss and numa_foreign statistics can be skewed | |
50 | heavily. In the current kernel implementation, if a process prefers a | |
51 | memoryless node (i.e. because it is running on one of its local CPU), the | |
52 | implementation actually treats one of the nearest nodes with memory as the | |
53 | preferred node. As a result, such allocation will not increase the numa_foreign | |
54 | counter on the memoryless node, and will skew the numa_hit, numa_miss and | |
55 | numa_foreign statistics of the nearest node. |