i am running a query in oracle 10 select A from B where C = D
B has millions of records and there is no index on C
The first time i run it it takes about 30
Clearing the caches to measure performance is possible but very unwieldy.
A very good measure for tracking achieved performance of tuning efforts is counting the number of read blocks during query execution. One of the easiest way to do this is using sqlplus with autotrace, like so:
set autotrace traceonly
outputs
...
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
1 consistent gets
0 physical reads
0 redo size
363 bytes sent via SQL*Net to client
364 bytes received via SQL*Net from client
4 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
The number of blocks read, be it from cache or from disk, is consistent gets
.
Another way is running the query with increased statistics i.e. with the hint gather_plan_statistics
and then looking at the query plan from the cursor cache:
auto autotrace off
set serveroutput off
select * from table(dbms_xplan.display_cursor(null,null,'typical allstats'));
The number of blocks read is output in column buffers
.
---------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows | Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers |
---------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 3 | | 1 (100)| | 3 |00:00:00.01 | 3 |
| 1 | SORT AGGREGATE | | 3 | 1 | | | 3 |00:00:00.01 | 3 |
| 2 | INDEX FULL SCAN| ABCDEF | 3 | 176 | 1 (0)| 00:00:01 | 528 |00:00:00.01 | 3 |
---------------------------------------------------------------------------------------------------------------------