From: Florian Weimer <fw@deneb.enyo.de>
To: mladen.gogala@vmsinfo.com
Cc: Greg Smith <greg@2ndquadrant.com>, Aaron Turner <synfinatic@gmail.com>,
	pgsql-performance <pgsql-performance@postgresql.org>
Subject: Re: large dataset with write vs read clients
References: <AANLkTimNJc=0ffgRO_gO6Vsq06=aqBYXZouhbv0kBe-i@mail.gmail.com>
	<4CAE3FDB.1040909@2ndquadrant.com> <87ocb3yu10.fsf@mid.deneb.enyo.de>
	<4CB0E006.8010905@vmsinfo.com>
Date: Sun, 10 Oct 2010 13:45:01 +0200
In-Reply-To: <4CB0E006.8010905@vmsinfo.com> (Mladen Gogala's message of "Sat,
	09 Oct 2010 17:35:02 -0400")
Message-ID: <871v7y2rwi.fsf@mid.deneb.enyo.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

* Mladen Gogala:

> I have a logical problem with asynchronous commit. The "commit"
> command should instruct the database to make the outcome of the
> transaction permanent. The application should wait to see whether the
> commit was successful or not. Asynchronous behavior in the commit
> statement breaks the ACID rules and should not be used in a RDBMS
> system.

That's a bit over the top.  It may make sense to use PostgreSQL even
if the file system doesn't guarantuee ACID by keeping multiple
checksummed copies of the database files.  Asynchronous commits offer
yet another trade-off here.

Some people use RDBMSs mostly for the *M* part, to get a consistent
administration experience across multiple applications.  And even with
asynchronous commits, PostgreSQL will maintain a consistent state of
the database.