Parallel processing for increaing the performance

Question

various ways of parallel processing in oracle especially using hints
Please let me knw if there exists any online documentation in understanding the concept

Answer 1

First of all: As a rule of thumb don't use hints. Hints make programs too unflexible. A hint may be good today, but might make things worse in future.
There are lots of documents available concerning parallel processing:
Just go to http://www.oracle.com/pls/db102/homepage?remark=tahiti and search for parallel (processing)
Due to my experience in 10g, enabling parallel processing might slow down processing extremely for regular tables. The reason are lots of waits in the coordination of the parallel processes.
If, however, you are using parallel processing for partitioned tables, parallel processing works excellent. In this case, take care to choose the partitioning criterion properly to be able to distribute processing.
If, for example, your queries / DMLs work on data corresponding to a certain time range, don't use the date field as partitioning criterion, since in this case parallel processing might work on just a single partition. Which again would result in massive waits for process coordination.
Choose another criterion to distribute the data to be accessed to at least <number of CPUs -1> partitions (one CPU is needed for the coordination process). Additionally consider to use parallel processing only in cases where large tables are involved. Compare this situation with writing a book: If you are planning to have some people writing a (technical) book consisting of just 10 pages, it wouldn't make any sense at all concerning time reduction. If, however, the book is planned to have 10 chapters, each chapter could be written by a different author. Reducing the resulting time to about 1/10 compared to a single author writing all chapters.
To enable parallel processing for a table use the following statement:
alter table <table name> parallel [<integer>];If you don't use the <integer> argument, the DB will choose the degree of parallelism, otherwise it is controlled by your <integer> value. Remember that you allways need a coordinator process, so don't choose integer to be larger than <number of CPUs minus 1>.
You can check the degree of parallelism by the degree column of user_/all_/dba_tables.
To do some timing tests, you also can force parallel dml/ddl/query for your current session.
ALTER SESSION FORCE PARALLEL DML/DDL/QUERY [<PARALLEL DEGREE>];