Thanks everyone for your responses. I don't think it's realistic to change our application infrastructure to use COPY from a stream at this point. It's good to know that multi-row-VALUES is good up into the thousands of rows (depending on various things, of course). That's a good enough answer for what I was looking for and we can revisit this if performance does start to hurt.
On 6/21/07, Andreas Kretschmer <[EMAIL PROTECTED]> wrote:
I guess you can obtain the same if you pack all INSERTs into one transaction.
Well the 20% gain I referred to was when all individual INSERTs were within one transaction. When each INSERT does its own commit, it's significantly slower. Steve