NUMCPUVPS and Fragments
Posted in 1997
On page 2-11 of the Informix Performance Guide Manual (Ver. 7.2), it states that "queries perform best when the number of scan threads is a multiple (or factor) of the number of CPU VPs". It then gives the example of having 6 CPU VPs, and a table with 10 fragments, and suggesting to reduce the VPs to 5. I was wondering if anyone could tell me, Why? The only explanation given is that "it produces an equal distribution of scan threads among CPU VPs". I would have though just the opposite. In the 1st case, it would process data from 6 fragments, then 4. Now if more scan threads return before it finished the 4 (or were already there when it finished the 6), now 2 more scan threads could begin being processed. But with only 5 VPs, it couldn't begin processing any from the 2nd batch until the 1st 10 were processed. That would seem to possibly make it slower (if the scan threads pulled data back faster than the VPs could process, although I would think that that case would be when there are a lot more scan threads/fragments than VPs). Thanks, Paul