J. ORIOL wrote:
I'm filling a 40 x 10 matrix of data (A string grid), that's 400
cells being calculated aproximately with 6 to 8 queries each one of
them. That's 4800 queries to fill the matrix. In the server, this
task delays 8 secs. but in the client are 40 secs.
What takes 40 seconds? If you're just running updates then the client
should receive acknowledgement.
The server slows down all its processes, sorry, does'nt "stop" tasks,
but in the practice, its the same, I think.
Certainly, I want to know if I have serious design errors or if I
have reached my HW limits; my hope was implementing functions in
plpgsql, but you guys said this will not improve the process time as
well.
You'll need to tell us something about your design. See my first reply
for what people will need.
As to whether your hardware is a limit - nobody can say without knowing
what you're doing with it.
Four obvious things leap out though:
1. Why does it take 6-8 queries to calculate one value?
2. What makes it necessary to calculate one cell at a time rather than
all at once?
3. Are you vacuuming enough, and how have you established this?
4. Why are you doing this in a database at all? For 4800 values I'd just
calculate them all in my application and then store the lot in one go.
--
Richard Huxton
Archonet Ltd
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
