We have two identical Oracle Exadata quarter racks each running a pair of database instances. My wet-finger-in-the-wind performance tests indicate that one is running at a quarter of the开发者_如何学Python speed of the other.
Is there a set of SQL queries that can be run with timers that would give a more scientific comparison of query performance between the two servers?
I'm aware that I could write a set of queries. I'm wondering if a semi-standard benchmark set already exists. Something like a few table creation queries followed by iterative insertion statements (lots of random data), index, constraint, trigger, function, proc & view creation scripts (preferably including materialised views) and then a bunch of complex queries that stretch the engine a bit with joins, trigger firing, function and proc calls, etc... while writing performance stats to some temp tables?
Anyone come across such a library/tool/script-set?
Try running the awrgrpt.sql for each rack and compare the output. Usually it is found here ${ORACLE_HOME}/rdbms/admin/awrgrpt.sql
This will give you a bunch of information like wait events, and top sql statistics. Should give you a clue as to why one has degraded performance over the other.
精彩评论