Processing Complex Events at the Large Hadron Collider (LHC)
Here is an excellent Google TechTalk by Dr. Majorie Shapiro (LBL & LHC ATLAS), June 18, 2007, on processing complex events at the LHC called, Supersymmetry, Extra Dimensions and the Origin of Mass: Exploring the Nature of the Universe Using PetaScale Data Analysis.
This is really what I would call “real-time CEP.” Bean collisions at the LHC happen every 25ns, and all the events produced are pipelined for decision processing to distinguish an interesting event from a non-interesting event.
The detection process uses a similar concept we outlined in our CEP functional model (based on levels in the JDL functional data-fusion model). Level 1 triggering preprocesses detected events with specialized detector hardware. Preselected Level 1 events are processed by computers looking for regions of interest (Level 2). These results are further correlated (Level 3) with results from the entire detector array (the whole complex event) and then sent to tape for archiving and future off-line processing.
Instead of explaining ATLAS at LHC (which I am not qualified to do), I recommend you watch the video yourself; it runs about one hour and 15 minutes. The video covers the computing processing required to process these complex events. Query processing, data storage requirments, raw bit stream processing, hierarchical event processing, event summary data, event tagging (classification for query purposes). event data statistics (amount, size), event generation simulation, distributed analysis, data mining and more event processing topics are covered.
Best quote in the talk:
“… how you do the analysis? - It ain’t SQL queries !! ” - Dr. Shapiro (0:48:44)