"Shridhar Daithankar" <shridhar_daithankar@persistent.co.in> writes:
> On 17 Sep 2003 at 0:16, Tom Lane wrote:
>> Well, yeah, they will. On a noncritical server, is that a sin? I mean,
>> if we offer fsync-off, it's not clear to me that offering WAL-off makes
>> the difference between venial and mortal sin.
> If somebopdy wants WAL effectively turned off, then can symlink WAL to
> a ramdisk that has a GB under the carpet. That would offer all the
> "benefits" of WAL being tunred off.
No, because the point of the proposal is to turn off WAL *temporarily*
during initial database load. Having to move WAL around and then back
again isn't simple, it isn't fool-proof, and it doesn't buy all of the
intended speed savings (the above might save some disk bandwidth but it
avoids none of the CPU expense associated with creating WAL entries).
regards, tom lane