Re: Import performance wierdness dicovered!?!

In a previous article, pape_at_plato.chemietechnik.uni-dortmund.de (David Pape) says:

>Hi all,
>
>I've been fooling around with the import program to find out if we can
>do imports more efficiently than at present, and have found the following
>to be true:
>
>When I import a single table of 3440 rows from an import file that is
>~21MB big (approx. 60 other tables data in the import file) and time it
>using /bin/time (Unix command, sorry DOSers), i get:
>
>81.3 real 38.2 user 11.3 sys
>
>i.e. 1.3 minutes real time to import the table.
>
>WHEN I export JUST the table of 3440 rows, and THEN import THAT file
>(now only 240 KBytes big), it took a mere 7.3 seconds of real time
>to import the table!!!!!
>
>What's up with that, eh?? Can this be explained by anyone out there?
>
>Does import have to search the entire file for extra extents? (I exported
>both files with compress=Y, so I wouldn't think this could be true.)

import has to scan the file to find the table you want to import. It then proceeds to import the table you want. (This scanning can take even longer if your file is on tape.)

The funny thing is that once your table has been imported it continues to scan the rest of the file until it reaches the end. I would guess that it may be looking for other tables of the same name (belonging to another user) Received on Fri Oct 07 1994 - 00:49:59 CET