Is there any optimization going on when doing query functions in expression analyses in PI AF?
In other products accessing PI data, like PI OLEDB, lots of effort has been given to create queries / functions which execute on the PI server instead of the client, in order to minimize the data being sent back and forth.
I had a PctGood(...) calculation evaluating the data quality for a tag the last 30 days. It was executed every hour as part of an element template with 2200 instances. Looking at the memory usage and the disk IO between the two (virtual) PI and AF servers it became pretty obvious that loads of data were being transported back to the AF server before being analyzed.