While trying to come up with some use cases for StreamInsight (SI) with real-time data, things boiled down to one question: what can you eventually do with StreamInsight that is hard to do now with PI tools. I ended up with only two categories:
-- Once SI is released, some GUI tools will show up on the market for on-line processing of real-time data. Such products could consist of GUI tools or visual frameworks to let you assemble data analysis strategies in a graphical way. So perhaps instead of creating ACE modules, one would have the possibility of creating plugins within such a framework. Think of a SQL Server Integration Services-like Visual Studio extension that acts on SI data.
-- Since data archived in PI goes through compression using a modified swinging door algorithm, some data mining strategies that are sensitive to such compression would prefer to process snapshots instead of archived data. However, apart from some purely academic papers, I could not find any application that woud suffer from using compressed data.