Hi,
You can also try to do VACUUM ANALYZE on all tables in mnogosearch database.
Or CLUSTER <index> ON <table> for indexes contain more one unique value.
Both should increase index seeks, but i not yet know what are better, possible VACUUM
ANALYZE.
On Wed, 7 Mar 2001 04:02:25 -0800 (PST)
Caffeinate The World <[EMAIL PROTECTED]> wrote:
CTW>
CTW> --- Alexander Barkov <[EMAIL PROTECTED]> wrote:
CTW> > Caffeinate The World wrote:
CTW> > >
CTW> > > mnogosearch-cvs, postgresql-cvs
CTW> > >
CTW> > > i just wanted to let everyone who uses postgresql with mnogosearch
CTW> > how
CTW> > > i got this incredible speed increase. I had about 500,000 URLs and
CTW> > when
CTW> > > i started indexing, it took for ever between each URL. per alex's
CTW> > > suggestion, i turned on debugging for sql.c and found where it took
CTW> > the
CTW> > > most time. a query like below took about 38 secs before i created
CTW> > the
CTW> > > index, now takes less than 1 sec between urls.
CTW> > >
CTW> > > [13289] SQL 0.26s: SELECT rec_id FROM url WHERE crc32=40757896 AND
CTW> > > status=200 AND docsize=10927
CTW> > >
CTW> > > i created the index:
CTW> > >
CTW> > > CREATE INDEX url_crc32_status_docsize ON url (crc32,status,docsize)
CTW> > >
CTW> > > perhaps this can be added to the create.txt file for postgresql.
CTW> > >
CTW> >
CTW> >
CTW> > Btw, does index on the only one crc32 field help?
CTW> > It uses less space and should be as faster as your one.
CTW>
CTW> i am waiting for this process to finish before trying just creating an
CTW> index for crc32. will update you.
--
___________________________________________
If you want to unsubscribe send "unsubscribe general"
to [EMAIL PROTECTED]
