> 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