Where all can the /+ PARALLEL / hint be used

hintsoracleoracle-11g-r2performance

It works with SELECT and UPDATE statements.

Are there some more statements with which it works?

I was able to use it with create, but am not sure if the hint was actually considered.

SQL> create /*+ PARALLEL */ table t1
( id number,
name varchar2(40));

Table created.

SQL>

Except INSERT and UPDATE, when all is the PARALLEL hint considered?

Best Answer

The Data Warehousing Guide has a chapter on parallel operations that goes over what operations can be parallelized.

You can't parallel-ize a simple CREATE TABLE statement (nor would you want to-- there wouldn't be enough work to be able to parcel it out to multiple worker threads). You can, however, parallel-ize a CREATE TABLE AS SELECT statement either by specifying the PARALLEL hint in the SELECT or by using the PARALLEL keyword in the CREATE.

Related Solutions

Best way to create multiple similar tables – CTAS vs INSERT

Use execute immediate to make creation of many very similar tables easier.

A little PL/SQL demo:

SQL> 
declare 
table_number number;
ctas_sql varchar(4000);
begin
  ctas_sql:='select sysdate as foo from dual';

  for table_number in 1..10
  loop
    execute immediate('create table foo'||table_number ||' as ('||ctas_sql||')');
  end loop;
end;
/
SQL> 
PL/SQL procedure successfully completed.

SQL> 
SQL> select table_name from user_tables;

TABLE_NAME
------------------------------
FOO1
FOO2
FOO3
FOO4
FOO5
FOO6
FOO7
FOO8
FOO9
FOO10

10 rows selected.

SQL>

The create table statement can contain anything you want - the above is just showing that the statement can be dynamic. So, in your case you'd have simple logic to alter the TEAM_ID and table name for each statement.

Bind variables will offer no gain unless you're executing the same single query multiple times (extremely simplified for the sake of this answer) - in this case you're executing many different queries that are performing large bulk inserts.

Creating tables with NOLOGGING is not advisable if this is a production system. Also be aware that not logging and /*+APPEND*/ can have backup/restore implications. This Ask Tom thread may be of particular interest.

To clarify, using CTAS is the way to go. You only have to parse the CTAS statement once to insert your millions of rows. If you use individual INSERT statements you have to parse each statement over & over again, which adds a massive overhead!

How to set the sequence to the last used id for all tables in Oracle

First you have to find out what the current sequence value is and then reset it. I would write a script which does the following tasks:

Get the max value from the table
update/recreate the sequence

The script could look like this:

declare
  maxval number;
begin
  select max(id) into maxval from <table_1>;
  execute immediate 'ALTER SEQUENCE <sequence_name_1> START WITH '|| maxval+1 ||';';

  select max(id) into maxval from <table_2>;
  execute immediate 'ALTER SEQUENCE <sequence_name_2> START WITH '|| maxval+1 ||';';
end;
/

I'm pretty sure there are better scripts but if you have to do this task frequently this might do the job without much development time.

Best Answer

Related Solutions

Best way to create multiple similar tables – CTAS vs INSERT

How to set the sequence to the last used id for all tables in Oracle

Related Question