hi all
i have been away from the oracle world for a few years, and recent growing
traffic caused our only oracle system running into its limit, so pulled me back
to oracle world…
i got two questions
1… whatis the realworld oracle concurrency limit? our 2way x86 server oracle
has reached 77k qps during peak time and ~225k user calls, running the system
at 70pct cpu(on flash card so no io wait)… i am afraid of even if we upgrade
to 4way/48core(now 2way/16core) server, internal concurrency limit(like library
cache mutex etc) will block system from 2x traffic… 5 years ago when i left
oracle world the most busy production system i knew was around 40k qps on a sun
t~3 server… wondering where there is many production system running at
100k~200k qps(a few thousand tps) in real world,if yes whatkind of hw
configirations?
2…i was curious about our system(11203 nom rac) with 225k user calls and 77k
executions… i know user call is usually higher than exec but this kind of gap
does not make sense to me…
we dont use plsql, no big sql fetching tons of rows(sql lio per exec was
just 9). a few hard parse(~60-100 per second) will cause more recursive call
but this does not explain.
anyone has idea?
or i should skip user calls and just read the executions?
thanks a lot
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
best regards
Mail: zhuchao@xxxxxxxxx
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
发自我的iPhone6
