Wed, 19 Sep 2007 03:28:11 +0000
The EPTS started its 3rd meeting as a lead up to the Gartner Event Processing Summit, alongside the BPM Summit, here in muggy Florida.
Day one went well [*1], with topics of Academia and Event Processing, and Application Design Patterns.
1. First off was Arno Jacobsen from the University of Toronto’s Middleware Systems Research Group , with a fascinating case study of a joint academic-industry project called PADRES, which exploits the brokers used in a pub-sub middleware system to act as agents that assist with event processing. Basically, the predicates involved with initial event processing are propogated to the broker(s) nearest the source(s) for that particular event, with the idea that network traffic can be reduced by not passing on any messages that are not required by the event processing rules. Neat, with a few caveats of course (like the complexity of combining your event processing rules and queries to predicates per event, then re-joining the events as necessary as they transit through the network). The prototype uses the JESS rules engine, and the entire solution is effectively doing content-based routing. [*2]
2. There was a discussion on academia, and how to get Colleges and Universities to educate Comp Sci undergrads and grads about Event Processing. This issue here is one of communication, and that often the CEP industry is “ahead” of the academia R&D establishments (notwithstanding Arno’s talk above). Solutions included vendors offering to do lectures (TIBCO is certainly happy to help here) and setting up an EPTS suggested curriculum.
3. Peter Niblett from IBM reported on the Dagstuhl seminar on CEP in Germany earlier this year, with a good attendance from industry as well as the (more usually represented) academia. Academic needs from industry to make progress included use cases with actual problem definitions as well as data. Suggested R&D areas in CEP were an event processing metalanguage (sounds ambitious!) and better modelling tools (for technical as well as business users).
4. Roy Shulte of Gartner then presented on Event Processing Patterns and Products - an excellent review of EP application patterns ending with an estimate of market growth. Interesting points here were that:
- you don’t need an “EP tool” to do EP (for example, consider manual event monitoring and transformation via eyeballs and a spreadsheet);
- a definition for “high-end” CEP which absolutely justified a “CEP tool” included 250+ events per second with 10 rules processed per event (although of course there was discussion on lower latency event problems with larger amounts of processing, and also the time-to-change of a CEP application, as other metrics);
- a comment on the emergence of EP as a solution for eXtreme Transaction Processing.
[*1] This is not a real-time, event-driven blog. There is a lack of wireless in the conference room. Some conference organisers argue that wireless access distracts attendees. I argue they lost that battle with the invention of the mobile phone, and have continued losing it ever since with the onset of Blackberries, cellular modems, etc.
[*2] Disclosure: TIBCO is of course a long-time provider of high-performance networking / messaging solutions. And usually these are used to provide events to CEP tools like TIBCO BusinessEvents. So agreement from us: pub-sub is good!