DEBS08(4) - Event-based Middleware
Posted by Paul Vincent
Continuing on with the Distributed Event Based Systems conference, its quite surprising that the presentations on event transmission (aka intelligent middleware, content-based pub-sub, clustering pub-sub nodes for performance, secure pub-sub etc etc) and of event processing (CEP, ESP, Bayes/rules/uncertainty/queries etc etc) are not in different “streams”, given they have different audiences. From a CEP perspective, middleware is certainly necessary (and for TIBCO, er, pretty relevant), but our interest is more on what to do with events, not worry about how they arrive.
Having said that, there is clearly a connection between “intelligent / optimal distribution of events” and “distributed (complex) event processing”. And “dynamic routing” implies “intelligent nodes” that are, in effect, limited event processing agents. So the concepts are connected, although the R&D goals may, today, differ (e.g. middleware: routing resilience, performance and scaleability; CEP: event processing scaleability, performance and resilience…).
A few observations were:
- In a presentation on gaming events, the definition of Event included an “obsolete” attribute, used to indicate an event whose existance is made “irrelevant” by some subsequent event(s). I’m not sure this is a generic attribute of all events, though - the event still happened, and could be correlated (for example to identify the speed at which such events become obscolete). So this is probably useful in some contexts, and not others.
- An event metamodel was presented by the University of Waterloo in Canada, which was much more comprehensive (/complex) that the definition we use in TIBCO BusinessEvents (for Event Processing). But their goal, to make “event” a first class entity, is shared by the OMG EMP team, with whom the Waterloo group should probably be collaborating.
Complex Event Processing (CEP), Meetings and events, Standards | 07 July 2008 |
1 Comment
Other Links to this Post
RSS feed for comments on this post. TrackBack URI
By P.J. Hinton, July 7, 2008 @ 05:36
I agree with you that on a theoretical level, CEP definitely shouldn’t be worried about how the events arrive, but I suspect in practice that the ability to get event data easily from disparate sources will be one of the critical hurdles to enterprise CEP adoption.
Has the group at U. Waterloo published anything on their proposed event metamodel? Although I haven’t been involved with the OMG EMP efforts since leaving Rhysome, I still like to keep a casual eye on how things are unfolding.
Cheers!
–
P.J. Hinton