On Tue, 2012-05-01 at 10:08 -0400, Robert Haas wrote:
> We've previously discussed the possible desirability of extending
> relations in larger increments, rather than one block at a time, for
> performance reasons. I attempted to determine how much performance we
> could possibly buy this way, and found that, as far as I can see, the
> answer is, basically, none.
Another point here is that with checksums, we will want to make sure
that zero pages can be treated as corrupt. That will probably involve
using the WAL for extension operations, and we'll want to mitigate that
cost somehow. Extending in larger chunks would probably be necessary.
There are some challenges there, but I think it's worth pursuing.
Regards,Jeff Davis