Hey Lothar,
physical I/O can cause (high) CPU usage too and only be caught/reflected by
getrusage(). Here is an example by Frits of such a behavior:
https://fritshoogland.wordpress.com/2013/11/24/when-the-oracle-wait-interface-isnt-enough/
In addition be aware that Oracle changed the accounting of direct path wait
events (Bug #7448407). The wait event can be "absorbed" but cause CPU
usage "in background" which is reflected by return value of getrusage().
Both scenarios could make sense here as Jonathan already pointed out the
multiblock I/O possibility.
Best Regards
Stefan Koehler
Freelance Oracle performance consultant and researcher
Homepage: http://www.soocs.de
Twitter: @OracleSK
Lothar Flatz <l.flatz@xxxxxxxxxx> hat am 2. Mai 2016 um 17:46 geschrieben:--
Hi Jonathan,
what also puzzles me is the CPU time. About the same number of buffer gets,
but 10 times more CPU. Can't remember where I have seen that before.
Regards
Lothar