Matthias Urlichs wrote:
> Hi,
>
> Chris:
> > Matthias Urlichs wrote:
> >
> > > Nonono, the 1000 read() calls are triggered by a simple INSERT or UPDATE
> > > call. They actually scan the pg_index table of the benchmark database.
> >
> > Does this only happen on the first call to INSERT/UPDATE after
> > connecting to the database, or does it happen with all subsequent calls
> > too?
> >
> All of them. Whatever the server is looking up here, it's _not_ cached.
>
Maybe shared buffer isn't so large as to keep all the(4.1M) pg_index pages.
So it would read pages from disk every time,
Unfortunately pg_index has no index to scan the index entries of a relation now.
However why is pg_index so large ?
Regards.
Hiroshi Inoue
Inoue@tpf.co.jp