Re: [Bizgres-general] WAL bypass for INSERT, UPDATE and

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: Hannu Krosing <hannu@skype.net>
Cc: Greg Stark <gsstark@mit.edu>, Qingqing Zhou <zhouqq@cs.toronto.edu>, pgsql-hackers@postgresql.org
Date: 2005-12-28T00:30:12Z
Lists: pgsql-hackers
Hannu Krosing <hannu@skype.net> writes:
> making one of WAL files (strategies) be /dev/null would almost get us
> non-logged writes, except for a little overhead in write() calls.
> fsync() on /dev/null should be instantaneous .

No, you really do want to push it up to a higher level than that.
The CPU and contention costs of generating a WAL record are nontrivial,
and doing so only to throw it away is pretty silly.  What's more,
pointing WAL at /dev/null would disable logging for the entire database
cluster, not just selected tables which is what people seem to be
asking for in this thread.  (No, I don't wish to deal with multiple
WAL output streams...)

			regards, tom lane