Demonstrating if Insertion is Faster in Partitioned or Non-Partitioned Table [message #65475] |
Thu, 30 September 2004 10:30 |
Shariq T
Messages: 29 Registered: February 2003
|
Junior Member |
|
|
I have posted a similar question on SQL forum but have not heard from them
I am working with a 600 million record table that has about 15 million records inserted monthly into it by an ETL tool that integrates the warehouse with a couple of billing systems. This table has a primary index and is NOT partitioned and of late some runs are taking well over 15 hours
After reading in several places how partitioning this table on month key will help with exactly with this sort of thing, I created a partitioned table in the test environment with populated it with identical data with identical colume
The partitioned table has local indexes whereas the non-partitioned one has global indexes. Each table is sized approximately 18 GB
I created a PL/SQL block that inserted dummy records into each of these tables. I committed after a block of 10,000 records was inserted into each table until 10 million records were inserted into the table.
The performance improvement/gain was only marginal for the partitioned table. Should this been happening? Can anyone offer a more elegant way or an example showing the benefits of using partition on this table over not using it
Looking forward to hearing from you
Shariq
|
|
|
|
Re: Demonstrating if Insertion is Faster in Partitioned or Non-Partitioned Table [message #65528 is a reply to message #65479] |
Fri, 15 October 2004 04:02 |
Mike
Messages: 417 Registered: September 1998
|
Senior Member |
|
|
Instead of inserting you data in the exiting partitioned table, create a new table and try to directly insert your data in there using a CREATE TABLE AS SELECT.
Once the table created, create an index for it, create a new empty partition on your partitioned table, execute a partition exchange.
In case you are using 9iR2 I recommand you to have a look at the new table/segment compression feature
|
|
|
|