High Waits on 'Db File Sequential Read' Due to Table Lookup Following Index Access (文档 ID 875472.1)
In this Document
APPLIES TO:Oracle Server - Enterprise Edition - Version 126.96.36.199 to 188.8.131.52 [Release 8.0.3 to 11.2]
Information in this document applies to any platform.
A query can wait on 'db file sequential read' for a long time even if the execution plan appears to be optimal.
This usually happens when the result set of index scan is large.
In most cases like this, the execution of the query waits on "TABLE ACCESS BY INDEX ROWID" much longer than on INDEX SCAN. It is because the random access to the table rows is much more expensive than index scan.
You can try one or more of the following.
1. Check if there is a better index or plan. You may need to re-design index configuration.
2. Try table full scan. Full scan often runs faster than Index scan, though its CBO cost is higher than the cost of Index scan.
3. Create a concatenated index to avoid the table access if the query has only a few columns of the table in SELECT and WHERE clauses.
NOTE : This can only be applied to SELECT for the table. If the query updates the table, this will not help.
4. Move the table into the tablespace with a larger block size. A larger block has more rows in it, so it may help to reduce block I/O.
And it will also be helpful to reorganize the table so that the index may have a smaller clustering factor.
5. Consider increasing buffer cache so that more table blocks can be cached. It is a good idea to use keep buffer pool if the table is frequently accessed.
6. Consider using IOT(Index Organized Table). IOT may reduce I/O because it stores data in a B*Tree index structure.
For example, if the column 'A' is the primiary key of BIG_TABLE, you can try to create IOT as follows.
create table BIG_TABLE (A number primary key, B char(2), C number, D varchar2(10))
7. Consider using parallel execution if there is sufficient free resource(CPU, memory) in the server.
This option does not reduce I/O. But, this may help to reduce execution time.
8. Bad disk I/O can be a reason. In this case, improve the I/O of devices where the table resides. This requires help from a system administrator.NOTE:34559.1 - WAITEVENT: "db file sequential read" Reference Note
NOTE:76374.1 - Multiple Buffer Pools