In MS SQL Server, it is well documented that the type of the primary key can have a dramatic effect on performance. Using narrower types and sequential values significantly improves insert performance when using SQL Server, which is why using integers instead of guids for surrogate keys tends to make sense.
However, I cannot find any references to the same discussion for Oracle. Someone asked me to look at a schema for them that uses varchar2(32)'s for all of the primary keys, which seems much wider than needed. Does this matter in Oracle?
Bonus points for anyone who can refer me to a good blog like Kimbery Tripp's but in Oracle world.
Best Answer
From what I know, it depends on whether your table is IOT (Microsoft calls it clustered table, a table with clustered index)or not . Historically, Oracle IOTs performed worse than its heap tables. That seems to change (I believe with Oracle 10), but there is still prejudice against IOT. So the article you mentioned can be applied to some extent if you use IOT in Oracle, but not for heap tables.
Don't misunderstand me, having a PK constraint supported by index on large varchar2 column has impact on performance, but if it's an Oracle heap table, it's the same as having an index on non-PK varchar2 column.