I'm in the process of implementing a "centralised" PI server that connects to the PI servers on site plus some Wonder InSQL server (we will hopefully replace these depending on the EA discussions). Most of our sites aren't blessed with huge network bandwidth. Therefore, I'm would like to estimate the amount of bandwidth that I would require for each connection. I've read through the PI-to-PI and RDBMS interface manuals. Based on these I would install the interfaces on the central PI server; the manuals indicate that there is a single call to retrieve the data and multiple calls to store the data into the archives. I can do a basic calculation based on the tag type (number of bits), number of tags and update frequency. However, this estimate can be out by a long way as it doesn't take into account system overheads. Therefore, does anyone know of formula for this type of calculation?