Do you want to know how much contention there is for CPU in your virtualised environment?
e.g. is VMware really scheduling your virtual machine when you need it? Have you ended up
with an EC2 instance that is being shared with some other CPU-heavy jobs?
am-i-scheduled sleeps for exactly one second, checks the time and then goes back to sleep
again. Every hour (configurable with the
--count option), it dumps a report on latency
and scheduling to stdout. Or, with the
--logfile option, to another file, or with
--syslog, it can write to syslog.
This is the kind of report it generates.
Reporting for 3 samples collected in the period between: - Thu Jul 6 14:57:32 2017 - Thu Jul 6 15:57:59 2017 Worst scheduling delay: 11.0209001 seconds -- Thu Jul 6 14:57:35 2017 Mean scheduling delay: 0.006195 seconds Standard deviation on delay: 0.0003 25% of the time, the delay was less than 0.000171 seconds Median scheduling delay was 0.002060 seconds 25% of the time, the delay was greater than 0.002090 seconds Times where we received no CPU allocation for half a second (or longer): 3