Re: Q: Imports
Date: 1995/06/22
Message-ID: <3sbc04$ens_at_bell.ukonline.co.uk>#1/1
Jonathan Lewis (Jonathan_at_jlcomp.demon.co.uk) wrote:
: In article <3s9dve$bei_at_bell.ukonline.co.uk>
: chris_at_perihelion.co.uk "Chris Selwyn" writes:
: : Could someone please tell me why creating hash clusters is soooooo slow?
: : I am trying to create a 150M cluster and so far it has taken ~3/4 hr
: : on a Alpha OSf/1 Oracle 7.1.4 system (and it's still going :-( ).
: :
: Rows in hash clusters are put into a block whose address is dictated by
: applying a function to values in the row, so each row inserted into your
: table is effectively going to a random position somewhere in the 150M.
: Unless you have a _very_ large SGA and file-system buffer, you are probably
: suffering from a very high rate of physical disk reads as Oracle is going
: to be flushing and re-reading blocks all the time.
: For a rough estimate of total time required, I'd bet on something like
: one-twentieth of a second per row.
The time appears to be going in the creation of the hash cluster : not in the population of it. Any ideas?
-- Chris Selwyn Senior Consultant Perihelion Software email: chris_at_perihelion.co.uk tel: +44 1749 344203Received on Thu Jun 22 1995 - 00:00:00 CEST