On Tue, Feb 9, 2010 at 10:42 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> A more important point is that it would be a nontrivial change, both as
>> to code and documentation, and it's too late for such in 9.0. So what
>> we ought to be confining the discussion to right now is what 9.0 should
>> print here.
>
> It's exactly as nontrivial as the proposed change in the other direction.
I think it's non-triviality lies in the user-visible effects. Ie, we
won't have a release cycle of experience with the change ourselves to
see if we really like the new behaviour and can live with it.
I'm not sure where I stand myself as I find the averaging quite
confusing myself. I'm not sure how confusing it would be to change
though.
The total i/o done is likely to be something you want to compare with
some other metric like output from iostat or knowledge of how much
bandwidth your i/o system can provide so it does seem like totals are
relevant.
I think I'm leaning -- slightly -- towards calling out the discrepancy
by naming it something like "Total Buffer Usage:". But I could be
convinced otherwise.
--
greg