Joins vs. Row chaining

Friends:

/* Oracle8i on Solaris 2.6 */

Let's say you've got a supertype table and five subtype tables off of it. The supertype is ~ 2K wide at most, and the subtypes are ~125 bytes wide. The database block size is 2K.

I'm considering collapsing the subtypes into the supertype, but I wonder which is GENERALLY more expensive I/O-wise: navigating a chained row, or doing a (two-table) join on unique-indexed columns. Is it simply a matter of adding up the number of I/O's needed to read the full row chain and comparing the results with the number of I/O's needed to read the index entries and the rows they point to? What am I not considering?

Thanks.

--Jeff

Thoughts Received on Thu Mar 30 2000 - 00:00:00 CST