James Owen’s “technical-biased” October Rules Fest returned to Dallas this year, and given it had a fair amount of coverage on Complex Event Processing, TIBCO contributed a paper titled “What’s Different about Rules in CEP“. ORF was held a few blocks East of a somewhat infamous Dallas location recognizable to many from a certain event 46 years ago.
I was very impressed by the content of ORF, although a minor quibble is that James persists in holding it as a separate event from BRForum and RuleML (both taking place the following week “up the road” in Vegas). James claims this is to avoid any marketing bias at a technical conference – fair enough, except that it’s the marketeers who pay the sponsorships that allow shows to take place.
From the Complex Event Processing perspective, coverage included:
- “Early Alert System at SouthWest Airlines” by Mark Sturdivant and Greg Barton – as it happens, a TIBCO BusinessEvents application – helping manage 540 aircraft over 68 airports. Interesting point here was the move from an event-centric design to a state + event centric design, and the classification of 2 main types of rule:
- rules maintaining the state of knowledge (concept structures) based on incoming events – a kind of “situation maintenance“.
- rules creating “alerts” responding to possible future states that are worthy of notification – or potential “incidents” – in other words “sense and respond“.
- “Temporal Reasoning – a requirement for CEP” by Edson Tirelli and Adam Mollenkopf, exploring the Drools’ rule engine’s new time expressions in their rule syntax and a CEP use case at Fedex (who incidentally are a very large TIBCO EMS customer).
- “A Survey on Complex Event Processing Models” – a very comprehensive coverage of the CEP language space by Charles Young, especially on possible interactions between stream-processing SQL engines and different parts of the Rete algorithm.
But CEP and attendent issues came up throughout the conference. Some interesting points were:
- Andrew Waterman presented on ecological solutions via educating farmers inadvertantly involved in agricultural over-exploitation and desertification in Central America, through the use of rule-based games. A new and likely increasingly populer term raised was for authentication of social network systems – Facebook authentication!
- David Holz from Grindwork presented on the use of declarative rules as a new generation software development methodology, describing how rules convert state to behavior. And “knowledge of state” is of course essential in CEP and a differentiator over stateless rule engines.
- Thomas Cooper of DEC XCON / XSEL 1980s expert system fame explicitly called out temporal / CEP models as being missing from most rule languages today. Interestingly, one of the extensions he’d added to the 70’s rule language OPS5 was to allow for effective dates on facts… probably the equivalent in the CEP world today is BusinessEvents’ concept history capability (storing the historic record of past values and their timespans). Thomas also lamented about rule system performance in multi-CPU systems (again, somewhat ameliorated by TIBCO BusinessEvents’ multi-threaded Rete and distributed agents capabilities).
- Dr Jacob Feldman presented on the merger of constraint solvers and inference rules – certainly an area to watch (and we are certainly very interested in this area of Business Optimization). A new JSR (JSR-331) has been set up by Jacob to standardize APIs for constraint solver execution, although surprisingly the CP vendor community is being somewhat slow to organize around this.
- Daniel Brookshier from UML vendor NoMagic presented on PRR as well as sponsored the event. Kudos to Daniel for implementing PRR *and* using Ruleby to provide a simple testing mechanism for UML users of PRR. This means there will be plenty to discuss at the next PRR standards meeting …
- FICO’s Carlos Serrano-Morales and Carole Ann Berlioz-Matignon (apart from winning the award for the longest names on the agenda) presented on the importance of measuring KPIs for rule “performance” (aka stateful monitoring of rule execution, another characteristic of BAM-type CEP systems) and business rules in the cloud and the importance of asynchronous events – another CEP characteristic. However, the FICO folk did not go as far as announcing any CEP offering.
- Luke Voss presented on a system of rulebased agents. Distributed agents are of course also a capability of TIBCO’s CEP system, although TIBCO’s agents do not comply with Luke’s requirement of “mobility”. Others may argue they do, though…
- Rete-inventor Charles Forgy presented the closing talk on parallel rulebases – something that probably benefits CEP more than simple decision services. Looked a bit like Map Reduce for rules to me! Charles also complained how far behind the Java JVM developers were in their failure to fully exploit multi-threading in Java.
For a somewhat less biased view of ORF’s sessions, check out Charles Young’s blog for days 2, 3, 4 and 5.
