First page Back Continue Last page Graphics

…Index Fast Full Scan

Notes:

The example demonstrates the index fast full scan. Prior to executing the query shown above, I had created a test table and index as follows:
create table t1 nologging as select * from all_objects order by object_id;
create unique index t1_object_id on t1 (object_id, object_name) nologging;
Note the addition of column OBJECT_NAME in the index. Though this was unnecessary (object_id was unique by itself), I added OBJECT_NAME to take advantage of the index fast full scan access path.
Note the inclusion of a SORT step in the execution plan because of the specification of the ORDER BY clause in the query.
Note that because a fast full index scan uses multi-block reads, the cost of this path divides the number of index blocks by the multi-block read count:
cost= mreadtim * (numblks / MBRC) / sreadtim
Refer to the supplied script index_ffs.sql for a copy of this demonstration.