Thread: Re: [pgsql-hackers] Daily digest v1.5568 (24 messages)
I wonder if this idea might be taken a little further, to allow read-only tablespaces? This would allow old partitions in very large databases to be kept on read-only media, and would allow normal backups to ignore this unchanging set of data. It also allows for certain specific optimisations for this type of data, as the MVCC rules are now relaxed. Inclusion of a row in a read-only index is now enough to guarantee the visibility of that row to all backends, and fetches based solely on the index now become possible. FWIW, Oracle does something like this. Just a thought. __ Marc On Mon, 2005-11-21 at 09:53 -0400, pgsql-hackers-owner@postgresql.org wrote: > Date: Sun, 20 Nov 2005 22:39:49 +0200 (EET) > From: Heikki Linnakangas <hlinnaka@iki.fi> > To: pgsql-hackers@postgresql.org > Subject: Data directory on read-only media > Message-ID: <Pine.OSF.4.61.0511201815450.399601@kosh.hut.fi> > > Hi, > > I've been thinking about running postgres from read-only media. It's > handy for creating demo CDs etc. I hacked together a patch that > allows > you to run Postgres without write permissions to the data directory. >
On Mon, 21 Nov 2005, Marc Munro wrote: > I wonder if this idea might be taken a little further, to allow > read-only tablespaces? > > This would allow old partitions in very large databases to be kept on > read-only media, and would allow normal backups to ignore this > unchanging set of data. I guess you could do that, but it's really quite a different problem. > It also allows for certain specific optimisations for this type of data, > as the MVCC rules are now relaxed. Inclusion of a row in a read-only > index is now enough to guarantee the visibility of that row to all > backends, and fetches based solely on the index now become possible. There's this TODO: > Allow data to be pulled directly from indexes > > Currently indexes do not have enough tuple visibility information to > allow data to be pulled from the index without also accessing the heap. > One way to allow this is to set a bit to index tuples to indicate if a > tuple is currently visible to all transactions when the first valid > heap lookup happens. This bit would have to be cleared when a heap > tuple is expired. That method doesn't require read-only tablespaces. - Heikki