Hi,

what's the effect of primary key field size on performance, as a rule
of thumb? Let's say that NUMBER(10) would probably be suffient size
for a primary key field. What's the order of magnitude of performance
degradation if NUMBER = NUMBER(38) is used instead?

Of course, too little information to be answered precisely, but can
the difference be considerable?

How does Oracle store NUMBERs internally - does it store NUMBER(9) in
32 bits and NUMBER(10) in 40 bits, NUMBER(10) in 64 bits or how?

janne_1976 (AT) hotmail (DOT) com (janne) wrote in message news:<6539de6f.0401070554.2a08b026 (AT) posting (DOT) google.com>...
Very good question, the difference is negligible the indexing leaf
'jumps' to the appropriate index. Boils down to a waste of space & how
fast your disk spindles are.

support (AT) DBtune (DOT) com
http://www.DBtune.com

What would you say about a "usual" database server with "usual" disks
and configuration (annoyingly vague, sorry - say for example some
basic Sun Fire UltraSparc box?) - is it fair enough to say that the
performance difference is probably negligible?

janne_1976 (AT) hotmail (DOT) com (janne) wrote in message news:<6539de6f.0401120613.795a9da7 (AT) posting (DOT) google.com>...
My interpretation of negligible time difference is not more than a few
seconds and so long as it's a 'usual' query, nothing fancy (or a trick
question) like a billion iterations. Each iteration will add on a
split second more, if this is the case then a huge difference can be
achieved. Please don't ask how long My original answer is based on
a simple 'usual' select statement...-Interesting no one else responded
here, this is a discussion I could be wrong, but I doubt it

FreeSupport (AT) DBtune (DOT) com
http://www.DBtune.com

janne_1976 (AT) hotmail (DOT) com (janne) wrote in message news:<6539de6f.0401070554.2a08b026 (AT) posting (DOT) google.com>...

You can find this part of the answer here

http://www.jlcomp.demon.co.uk/faq/num_size.html

Regards

Paul