How do you update very large tables in Oracle (tens or hundreds of millions of records)? You don’t!
Ok, this isn’t anything new and it’s not rocket science, but maybe it’s a technique that you haven’t used before. If you’ve tried to update millions or hundreds of millions of records in a table, you know that this can be an extremely time consuming process that ties up resources (and, may end up crashing anyway). Here’s a simple solution that avoids using an update statement (and, all the redo associated with it):
Let’s say you have a table with 50 million records and you need to update one of the columns using by performing a calculation or applying a function.
At a high level…
- Create a new table by selecting everything from the old table and performing the calculation in the select clause
Create table TABLE_NEW nologging as select <perform calculation here> from TABLE_OLD;
- Apply all constraints, indexes, grants, etc. to the new table
- Drop the old table
- Rename TABLE_NEW to TABLE_OLD;
Note the nologging option bypasses generating any redo and will result in significant performance improvement.
That’s it!
Of course there are other options like partition swapping or writing a stored procedure (never do procedurally what can be done in a single sql statement). But, I like this because it is safe and clean.
