Our company is thinking about PI Integrator for BA for a few use cases, namely 1. Operational Reporting, 2. System Integration (feeding data to commercial apps), 3. Advanced Analytics (steaming analytics).
We have all PI components on-premise, and plan to leverage cloud (Azure) services as much as possible.
For Advanced Analytics, the architecture is "PI DA -> AF -> PI Integrator (advanced edition) -> Azure Kafka -> Azure Data Lake -> Azure Databricks -> Visualizations".
However, we do not have a clear architecture in mind for the other two use cases.
For Operational Reporting, I'm thinking "PI DA -> AF -> PI Integrator -> PI Views (SQL Server) -> PowerBI".
- Do I have to introduce SQL queries/procedures on top of PI Views if further data manipulation is needed? If so, how easy to query PI Views in SQL Server?
- Will performance be an issue over long time since PI Views is actually staging PI data in SQL Server? I do not think it's a good idea to publish high-frequency data to PI Views because of performance.
- What about "PI DA -> AF -> PI Integrator -> Azure SQL Data Warehouse -> PowerBI"? SQL Warehouse is more capable of storing data, plus more options for further data manipulations.
For System Integration, what would be a reasonable architecture for feeding PI data to other applications (one-way)? I'm thinking PI Integrator publishes data to PI Views or SQL Data Warehouse, and other applications would grab from there.
Again, I'm not sure which one is a better option, PI View, Azure Data Lake, Azure SQL Data Warehouse.
Overall, we would like to have less components in architecture to support these 3 use cases. Your comments and recommendations are very appreciated.