Using joinfiles w/ large databases, referential integrity

I do a lot of database development, but haven't worked with DB2 on AS/400 too much since most of my previous work has been in Oracle and MS-related stuff on Windows.

We are doing some work now for a company which gets their data from a different company which has a large database on DB2 running on AS/400 or some other IBM mainframe (I don't know how large, but it is large .. presumably something in the terabyte range).

Everything is in flat files, no FK constraints, PK constraints are only on paper, nothing is journalled, etc. All referential integrity and business rules are apparently implemented by server procedures, and the data is only in 1st normal form (at best).

Since we only read their data, it doesn't really matter to us as long as they keep it consistent. Just out of curiosity, though, I asked one of their developers why there are no constraints in the database and the response was that it is apparently quite painful to rebuild the joinfiles (i.e. constraints) with a large database on AS/400 if they ever "break" (whatever that means, "painful" meaning a downtime of several days).

So my question to you is: How many people work with such large databases and still use constraints with referential integrity enforced through joinfiles? Is there another possibility? Or is it just too impractical above a certain size? What limits are considered too large for DB engine-level RI in the AS/400 and mainframe world?

As I had always been taught to normalize and use sufficient constraints with RI, I must admit that I had trouble getting used to the idea that many serious applications don't use them at all.

Bob Hairgrove
rhairgroveNoSpam_at_Pleasebigfoot.com Received on Sat Sep 21 2002 - 12:36:26 CEST