Hi Dave,
I don't think I can improve this by tuning. I've already played with the indexes, changed the shared buffers, etc. I've got to the amount of 3GB for shared buffers, wich is close to the best results I've got grom pgbench. The machine is a dedicated IBM Server with 16GB RAM and two Quad Cores (Linux sees it as 8 processors). If there's any way to improve this query with server tuning, it's beyond my knowledge to do it.
That's why I've thought about tuning the query, wich seems to be a complex task. Do you think there's any way it cab be tuned? I couldn't find any looking at the data model.
