I ever been in doubt for how long should we collect system stats. The
majority of our system don´t have an OLTP vs DSS period. So the common
approach for collect system stats (and store the for later reuse) for OLTP
and DSS periods sounds no reasonable to me.
My current approach is collect for a full week period.
Can others please add his/her experiences on this subject?
*--*
*Att*
*Luis Santos*
2016-09-02 4:29 GMT-03:00 Jonathan Lewis <jonathan@xxxxxxxxxxxxxxxxxx>:
A simple explanation would be that you gathered system stats over a period
(start/stop option) in which no parallel queries occurred. Those figures
allow the optimizer to calculate the maximum effective degree of
parallelism by comparing the maximum throughput the QC can take compared to
the average throughput a PX slave can produce.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
@jloracle
------------------------------
*From:* oracle-l-bounce@xxxxxxxxxxxxx [oracle-l-bounce@xxxxxxxxxxxxx] on
behalf of Deepak Sharma [dmarc-noreply@xxxxxxxxxxxxx]
*Sent:* 02 September 2016 02:46
*To:* Oracle-L Freelists
*Subject:* System Stats - MAXTHR and SLAVETHR
A 10053 trace showed these as '-1 bytes/sec'.
What does that mean?
