For Linux systems, the load average reported via commands such as `uptime` is "a measure of the average number of processes that are either in a runnable or uninterruptable state".
To be more concrete, the following scenarios all contribute to the load average metric:
- A unit of work is on-CPU and being executed
- A unit of work is off-CPU in the schedule queue waiting for it's turn to be on-CPU
- A unit of work is waiting for an I/O resource to become available (disk/network)
Therefore, it is useful to keep in mind that load is not strictly a measure of CPU "busyness" since for Linux systems I/O plays a part in the value that is calculated.
No comments:
Post a Comment