Documentation/accounting/psi.rst

4 PSI - Pressure Stall Information
14 either play it safe and under-utilize their hardware resources, or
23 scarcity aids users in sizing workloads to hardware--or provisioning
34 Pressure interface
37 Pressure information for each resource is exported through the
38 respective file in /proc/pressure/ -- cpu, memory, and io.
48 The "full" line indicates the share of time in which all non-idle
67 Monitoring for pressure thresholds
71 pressure exceeds certain thresholds.
78 /proc/pressure/ representing the resource to be monitored and write the
85 For example writing "some 150000 1000000" into /proc/pressure/memory
87 1sec time window. Writing "full 50000 1000000" into /proc/pressure/io
102 The kernel accepts window sizes ranging from 500ms to 10s, therefore min
103 monitoring update interval is 50ms and max is 1s. Min limit is set to
116 Notifications to the userspace are rate-limited to one per tracking window.
118 The trigger will de-register when the file descriptor used to define the
142 	fds.fd = open("/proc/pressure/memory", O_RDWR | O_NONBLOCK);
144 		printf("/proc/pressure/memory open error: %s\n",
151 		printf("/proc/pressure/memory write error: %s\n",
158 		n = poll(&fds, 1, -1);
182 mounted, pressure stall information is also tracked for tasks grouped
184 cpu.pressure, memory.pressure, and io.pressure files; the format is
185 the same as the /proc/pressure/ files.
187 Per-cgroup psi monitors can be specified and used the same way as
188 system-wide ones.