Re: Oracle vs. DB2 UDB

On 7/14/06, Allen, Brandon <Brandon.Allen@xxxxxxxxxxx> wrote: 

 I don't know anything about DB2, but you can only co-locate to satisfy
one sort criteria/index.  You can do this manually in Oracle by reorganizing
a table and ordering it to match a given index (and thereby greatly increase
the clustering factor for *that* index, but possibly blowing it for the
others), or you can use Oracle's IOTs (index organized tables) if
appropriate - that's what SQL server does if I understand correctly.  Been a
while since I worked on SQL server, but IIRC, the default for every table is
for it to be index organized according to it's primary key, rather than just
a heap like the default in Oracle.


actually the default bahaviour of sql server is heap tables just as in
Oracle. The 'conventional wisdom' however (and so the default behaviour of
sql server developers) is that every table should have a clustered index
(the feature you are describing) and that that should be the primary key
where there is one. Personally I didn't go much for that 'wisdom' myself. As
an aside my observation was that there was one very good side effect and
that was that people from a sqlserver development background actually tended
to create primary keys and foreign keys more often than people from and
Oracle background. Not necessarily the right ones (and nearly always
artificial ones) but none the less relational databases with relationships
in them. :(

--
Niall Litchfield
Oracle DBA
http://www.orawin.info

Other related posts: