As the PI system contiues to become more mission critical, we are now considering implementation of a PI Data Archive Collective during our Event Frame conversion activities. The goal would be to never have downtime and boost performance.
I have several questions for those of you who are are very familiar with implementation of a PI Data Archive Collective. This includes both customers and OSI technical resources:
1) Can anyone give me a feel for the extra overhead, maintenance, and increased complexity after having implemented a PI Data Archive Collective?
2) What sources of information\documentation exists to aid in implementation of a Data Archive collective other than what exists in the live library under PI Server 2017? Are there any white papers, best practices, or more detailed information how to walk the system over including how the new collective member would interact with the AF? and how that is configured?
3) I'm concerned information I've read about not all data getting replicated. We have processes that write data into the archives via PI ACE=>C# Code=>PI OLEDB=>PI Data Archive (time,tag,value). It sounds like that would not be replicated. I'm not sure if making PI OLEDB dependent on the PI Buffer subsystem would have an impact when writing directly into the Data Archives using PI OLEDB.
4) What is the implication of have data archives (primary and secondary) not matching exactly due to item 3 or similar? If a client doesn't know what it's connected due to the nature of a collective then will that result in seemingly random results depending on where the client actually gets connected? How can that be dealt with?
We have been toying with the idea of a PI Data Archive Collective for years but there is always some gotcha that stops us. We have a batch process, make heavy use of the PI Batch Database the limitation until now was that batch data could not be replicated. That limitation will no longer exist as we convert to Event Frames but now I'm wondering if we are still in a position where implementing an HA PI Data Archive is not possible due to the other limitation listed above.
Please feel free to expound with general helpful advice on PI Data Archive Collectives. If you have implemented one, are you glad you did?