Tricks for super-fast batch updates with CTAS

This is a secret that is not taught at Oracle University, a trick known to all DBAs who spend late nights, weekends and holidays performing database maintenance during tight windows of opportunity.

When you are updating the majority of rows in a table, using Create Table as Select (CTAS) is often more efficient performance than a standard update. The CTAS method employs this technique:

• The database is offline and a full backup will be done after the table is modified.

• All indexes and constraints are dropped and rebuilt after the table update.

• The decode clause is used to change the data when the table is copied. (e.g. select decode (status,'new','old',status from mytab;)

• The CTAS method uses the nologging and parallel method to speed up the table copy.

• After the table is copied and re-named, all indexes and constraints are re-created in nologging and parallel mode.

The CTAS method is designed for massive updates, and it is used by DBAs during batch windows when the database is in maintenance mode and no other operations are being done against the target table.

The CTAS method is fastest when the vast majority of the table rows are effected. As a general rule of thumb, any update that effects more than half the table rows may see faster performance with a CTAS update.

For example, assume that the following update changes 75% of our table rows:

update 
   mytab 
set 
   status = 'new' 
where 
   status = 'old';

In this case, a parallelized CTAS may perform far faster (Note: Make sure that you have an SMP server before using the parallel degree option):

create table 
   new_mytab 
NOLOGGING as
select /*+ full parallel(mytab,35)*/ 
   decode (status,'new','old',status,
   col2, col3, col4
from mytab;

-- rebuild indexes, triggers and constraints to new_mytab

rename mytab     to bkup_mytab;

rename new_mytab to mytab;

In cases where the updates are part of a scheduled weekend batch, this technique can reduce a large table update dramatically saving precious hours during scheduled maintenance windows.

Tip!  Gathering stats on a CTAS is no longer necessary in 12c, provided the CTAS statement is issued by a non-SYS user.  Prior to 12c you would need to analyze the table (dbms_stats.gather_table_stats) to get the metadata for the SQL optimizer:

EXEC DBMS_STATS.gather_table_stats('SCOTT', 'EMP1');