A newer version of this document is available. Customers should click here to go to the newest version.
Recording Operating System Counters
Similar to recording of process specific counters, Intel® Trace Collector can record operating system counters, which provide information about a node. In contrast to the process specific counters, OS counters are sampled very infrequently by one background thread per node and thus the overhead is very low. The amount of trace data also increases insignificantly.
By default, recording of OS counters is disabled. To enable it, set the configuration option:
COUNTER <counter_name> ON
|KB/s||Read/write disk IO (any disk in the node).|
|KB/s||Read/write network IO (any system interface). This might not include the MPI transport layer.|
|percent||Average percentage of CPU time of all CPUs spent in idle mode.|
|percent||Average percentage of CPU time of all CPUs spent in system code.|
|percent||Average percentage of CPU time of all CPUs spent in user code.|
You can change the delay between recording the current counter values with the configuration option OS-COUNTER-DELAY (by default, 1 second). CPU utilization is calculated by the OS with sampling, therefore a smaller value does not necessarily provide more detailed information. Increasing it could reduce the overhead further, but only slightly because the overhead is hardly measurable already.
These OS counters appear in the trace as normal counters which apply to all processes running on a node. To view the counters in Intel Trace Collector GUI, use Counter Timeline.
Did you find the information on this page useful?