PI Data Transactionalisation

The PI historian is designed to be very good at storage and retrieval of continuous data. Its highly-efficient compression mechanisms have been designed around the physical constraints of sensors and process measurement. This approach offers significant advantages over relational databases in terms of data-storage volumes and access speed.

However, there inevitably arise certain circumstances when the relational model has some advantages. A KSC client required a means of "chopping up" the continuous data streams in PI into discrete chunks, in order to create relationships between event-bounded summary figures and external operational events.

The system can for example, monitor the grade of crude oil into a distillation column. When the grade changes, it will generate a totalized throughput figure for the grade just ended. Thus, the aggregates are event-bounded, as opposed to the standard time-bounded aggregates that ship out-of-the box.

The system was designed with a user interface that allows event boundary rules to be defined, along with the corresponding aggregation mechanisms. The user can also specify where the output records go - to a web service, file system or relational database, from whence the newly-transactionalised data can be consumed by other business systems.