> >> We could retarget to try to stay under SHMMAX=4M, which I think is
> >> the next boundary that's significant in terms of real-world platforms
> >> (isn't that the default SHMMAX on some BSDen?). That would allow us
> >> 350 or so shared_buffers, which is better, but still not really a
> >> serious choice for production work.
>
> > What is a serious choice for production work?
>
> Well, as I commented later in that mail, I feel that 1000 buffers is
> a reasonable choice --- but I have to admit that I have no hard data
> to back up that feeling. Perhaps we should take this to the
> pgsql-perform list and argue about reasonable choices.
Damn. Another list I have to subscribe to!
The results I just posted indicate that 1000 buffers is really quite bad
performance comaped to 4000, perhaps up to 100 TPS for selects and 30 TPS
for TPC-B.
Still, that 1000 is in itself vastly better than 64!!
Chris