> I am just suggesting that instead of flushing the log on every
> transaction end, just do it every X seconds.

Or maybe more practical is, when the log buffer fills. 
And of course during checkpoints.

Andreas

Reply via email to