RE: Berkeley DB... - Mailing list pgsql-hackers

From Andreas Zeugswetter
Subject RE: Berkeley DB...
Date
Msg-id 00052809330604.00145@zeus
Whole thread Raw
In response to RE: Berkeley DB...  ("Mikheev, Vadim" <vmikheev@SECTORBASE.COM>)
List pgsql-hackers
On Fri, 26 May 2000, Mikheev, Vadim wrote:
> > > Well, I've dropped index but INSERTs still take 70 sec and 
> > > COPY just 1sec -:(((
> > 
> > Well, for those that have fsync turned off we could actually 
> > avoid most of the writes, could'nt we ? Just leave the page
> > marked dirty. We would only need to write each new page once.
> > The problem as I see it is, that we don't have a good place 
> > where the writes would actually be done. Now they are obviously
> > done after each insert.
> 
> I've run test without fsync and with all inserts in *single*
> transaction - there should be no write after each insert...

Yes, but if you don't do the inserts in one big transaction
and don't issue transaction statements ( no begin or commit )
then you get the behavior I described.

Andreas


pgsql-hackers by date:

Previous
From: Andreas Zeugswetter
Date:
Subject: Re: Proposal for enhancements of privilege system
Next
From: "Mitch Vincent"
Date:
Subject: Full text indexing preformance! (long)