Re: Oracle CPU Time Used, vs CPU Time Available (CPU Capacity)

Hey Chris,

> Given # of Cores _at_ xClockSpeed, how much time is available to the system?

Well unfortunately or better said luckily this is not how it works nowadays. Even the example from Peter ("Over a 60 second period, You have 1 CPU, 2 transactions arrive at the same time, They each use 15 seconds of CPU. Because the transaction arrive at the same time and you can only service one at a time, response time will be over 30 seconds for at least one of the transactions.") is not "true" as the normal CFS CPU scheduling (SCHED_NORMAL) which is used for Oracle on Linux for example - ignoring vktm process (hidden parameter "_highest_priority_processes") - works differently otherwise Oracle would may be able to block the whole CPUs (real-time schedule but still controllable by sched_rt_runtime_us & sched_rt_period_us or schedule enhancements in Linux kernel >= version 3.14).

This whole topic would go very deep into CPU scheduling and CPU architecture - so let's try to keep it at a high-level.

1. Oracle's CPU accounting is not very precise (from a CPU's point of view) - only ±10.000 microseconds due to getrusage() and scheduler interrupt.
2. The CPU time is split into CPU timeslices by CFS. The timeslices for a given task are calculated just before it is scheduled for execution. Its length is variable (dynamic).
3. The xClockSpeed of each core determines how much CPU cycles per second are possible - so basically you can also calculate the amount of CPU cycles for your specific assigned timeslice.

The problem is to determine how much timeslices are assigned to your processes and even then you don't know how many CPU cycles you need for each instruction as it also can be stalled (e.g. RAM access, etc). Of course you can measure all of these things (e.g. with perf) but this is mainly just for researching and hacking fun :-)

Best Regards
Stefan Koehler

Independent Oracle performance consultant and researcher Website: http://www.soocs.de
Twitter: _at_OracleSK

> Chris Taylor <christopherdtaylor1994_at_gmail.com> hat am 13. September 2017 um 21:42 geschrieben:
>
> Many of us (if not all) are familiar with Oracle counting CPU Time used in various metrics right?
>
> So how can we calculate the time "available"?   Is there some way to say:
>
> Given # of Cores _at_ xClockSpeed, how much time is available to the system?
>
> If we could calculate that, we could say "this process used xCPUMinutes of yCPUCapacity".  
>
> Just curious if anyone knew a way to do that?
>
> Chris

--
http://www.freelists.org/webpage/oracle-l