1, logical reading or physical reading?The actual execution plan of the query statement.F5 expected execution plan.-How to produce the actual execution plan??. --session Collection instructions.
Workload Repository Report for AWRElapsect:dbtime. Numerator/Denominator ease:->1.
--top FOREGROUND EVENTS by Total WAIT time--sql ORDERED by GETS the number of disk buffers--sql ORDERED by VERSION COUNT child cursors are arranged. Same statement < Use FALSE hint to complete, reduce child cursor polling. (Hard parsing > Soft parsing, 2.RAC)--SELECT * from TABLE (Dbms_xplan. Display_cursor (sqlid, child cursor number <null represents all, which data < personal preferences in the execution plan;.--e-initial estimate.--a-actual logic count (statistics session)
--sample:dev,prd,qas. Data volume ↑, CBO way optimization, go index way different.--Background: Table connection: 4. substr, | |--buffers: Number of disk cache executions. Dependency: Parameters: 1) Number of blocks per read, 2) actual environment. Example: Parameters: 16. Actual storage block, distribution, continuous reading of a. 8,18. Read 8,16,2.Independent of tablespace logical size, actual storage, internal mode control, fragmentation.--Method: Shorten the index, rowid→ the actual value cost. Increase the Index organization table query. (The query result column is placed in the index, and the number of queries is not cost-effective.) Cbo. For query excute more times, the effect is obvious). Reduce get.
--average per row occupancy, tuple storage size. property is not occupied. Estimation method: Average row data occupies length * How many rows. Analogue data, table data acquisition, average length. < Table Analysis >--dbms_stats.gather_table_stats data statistics. Automatic trigger mechanism + manual statistics. Schema-based statistics, cost↑, table-based statistics.The Art of Oracle Performance Diagnostics.
From for notes (Wiz)
DBA Consultant Training Content @20141230
