Re: Index for very small tables needed ?

Hi Jonathan!

Jonathan Lewis writes:

> Simple tests show that the "break point" from an indexed path to a
> scan path can occur at dramatically different volumes of data because
> of the way that the MULTIPLE cheap cost of finding and testing lots of
> rows compares with the rare, but expensive cost of a block access.

This is something I have been thinking about for a long time now. What *exactly* causes the CPU time for accessing maybe 200 rows scattered over 3 blocks which are already cached in memory?

Does Oracle step through the normal access path (hashing the block address, scanning the cache buffers chain, ...) for every single row or what makes it so expensive in terms of CPU time?

Regards,
Stefan Received on Wed Jan 02 2002 - 11:52:23 CST