Re: Locally Managed Tablespaces ... again!!!

"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in news:dDJU9.23562$jM5.62446_at_newsfeeds.bigpond.com and I quote:

> you are going to propose that we need to separate one segment away from
> another because of multi-user contention, then you might as well argue
> that every table needs to be in a separate tablespace, too.

Er....
Don't look now, but I've got a DB2 tuning book next to my bed that recommends *PRECISELY* that!
<sigh>

> It's a question of analysing I/O contention issues, not "These are
> indexes so they go there; and these are tables so they go here".

Absolutely. I/O contention needs to be addressed as such, not as the conventional wisdom. Measure where is the hot spot, split the load *right there*. Couldn't care less if that is table/index boundary, table extent/partition boundary, whatever. That's where and why LVMs in Unix become so useful.

-- 
Cheers
Nuno Souto
nsouto_at_optusnet.com.au.nospam