What we want to show
- system can really handle real-time data: run some workflows, measure latency of events from end of job (when data is transferred) to when it is inserted in the DB
- we have some useful ways of visualizing and digging into the workflows
- we can do some real-time analysis
- if we could put those derived events "back on the bus" (BoB-bing for derived data?), then other tools could do arbitrarily cool things with them. For instance, the workflow browser could use them as jumping-off points. Or some other analysis program could count the number of anomalies in a time-window and send email when it exceeded a threshhold.