2 of 2 people found this helpful
With regards to your questions:
1) It is not required to configure PI Interface on a separate node from the PI Data Archive however it provides an additional layer of protection when it comes to data loss. With your PI Interface on another node, you can configure buffering which will allow you to store the data locally on the interface in the event of your PI Data Archive going down (offline or unavailable for other reasons). If you have your interfaces on the same node as your PI Data Archive, if something happens to that machine, your interface will also be offline and thus you will lose data during that period. Having separate nodes also makes it easier to continue data collection during system/maintenance, upgrades, etc. Some additional resources that shed more light onto this are topics like data loss vs data availability as well as this PI Interface Node Architecture - White Paper
2) The "buffering" behavior that occurs between the snapshot are archive (known as the event queue) is automatic; you do not need to configure it aside from specifying the location for your event queue files which is done during your PI Data Archive installation.
3) Compression has default values but can be changed on a point-by-point basis. Depending on what the data is, how much granularity you need, and the precision of the instruments/sensors that are collecting the data, you may want to tighten compression (retain more values) or loosen it (store fewer values that still capture the overall shape\trend of the data within the limits of your needs).