RE: separate tablespaces for tables and indexes

"All of your arguments about 'indexes and their tables were read
anyway', to me, only really holds up when, as you say, there is only one
user on the machine."

When there are a hundred users on a machine and your top 10 busiest
segments happen to be table segments you have accomplished nada by
splitting the index segments.  So multiuser system still does not
matter.  Why not have a rule split all tables > 200MB on separate disk,
split tables starting A-M on Disk1 and N-Z on Disk2, you see these rules
are all arbitrary just like the split index/tables argument.

Tables are also accessed sequentially (just like indexes) so you may
have table read, table read, index read, table read.  It doesn't matter
if that is one session or hundreds.

Other related posts: