Re: [HACKERS] Postgres Speed or lack thereof

Hannu Krosing <hannu(at)trust(dot)ee> writes:
> I am sure that we need to change the protocol sometime soon, if we
> want to increase the performance.

The immediate problem is just that we've given back performance that we
had a week ago. stdio-based libpq didn't have this problem, because it
dealt with the kernel a bufferload at a time not a character at a time.

Also, while the fe/be protocol is certainly a bit crufty, I'm not
convinced that a well-designed frontend is going to suffer any real
performance problem from parsing the protocol. The trick again is to
read into a buffer and parse from the buffer. The last time I profiled
frontend libpq, it turned out to be spending all its time in memory
allocation not parsing. (Which I fixed, btw.)

> And of course we need some sort of CLI that can do prepared statements
> and that can use binary fields (not only the file interface).

> Currently we have some of it, bu only in SPI.

SPI? I missed something ... what's that?

regards, tom lane