Business Rules Forum 2008: Business Decisions from Real Time Events

[iBot]

vincent
10-29-2008 05:28 AM
Business Decisions from Real Time Events was TIBCO’s contribution to introduce event-based rules and decisions to the business rules audience. This covered:

1. Conventional IT systems, even SOA and BPM, even with predictive analytics and decision services, are still replicating the pneumatic-tube driven Victorian-era office - one message at a time, point to point. This works for things like human event processing / workflow, and automated manual systems in a predominantly manual business system, but is not in any way exploiting all the potential of IT… which is where things like EDA and CEP come into play.
2. There are multiple ways of processing abstract events (queries, rules, state models, etc). By themselves, these are not “new technologies”. But they are being applied and combined in new ways for continuous event processing.
3. For terminology, there has been some good work done by the cross-industry EPTS body (with other areas of cooperation planned too). [*1]
4. The CEP space appears to be (per the analysts) growing faster than even the “conventional” (BPM-oriented) Business Rule space. This statement may appear to be “hype” if you don’t believe analysts. It may simply be a reflection on EDA adoption vs WS-based SOA. Or the natural evolution and merger of multiple fields and best practices like decision management and complex events.

Naturally, as this was a rules conference, I focused on rule-driven CEP rather than query-based stream processing, and my baseline CEP architecture was that provided by TIBCO BusinessEvents (arguably a superset of other CEP offerings anyway). As usual, the Q&A was interesting (although I may have missed reporting a few below) and reflected the technical audience:

• Was the Rete algorithm, as used by conventional BRE technologies, applicable also to CEP?
  Answer: Yes. Currently only TIBCO among the mainstream CEP vendors supports this technique, but others are working on it.
• For traceability, do you have to record/log (and store) *every* event in a CEP system?
  Answer: It depends on the event, whether you need to for compliance reasons, or can simply aggregate its metadata in an abstract event… so this is very application and event-dependent. And of course, storing everything might have a performance cost at low latency levels.
• What is the performance limit for CEP systems, in terms of events per second, and what happens when you exceed that?
  Answer: This is similar to the question asked of conventional rule engine vendors: how may rules can you fire per second. The technical answer is “it depends on what the rules are doing”. The business answer is “enough”. For event-driven systems the bottlenecks might be the message bus and network interface as well as the processor. But that is also why you often find the capability for distributed load balancing in CEP technologies: add another box.
• Are there any standard benchmarks like TPC for CEP?
  Answer: Not yet, although the event stream vendors have done some work with STAC and there is an interesting EPTS initiative to work in this area.
• How do CEP event stores relate to the interesting growth in new data storage technologies?
  Answer: the need for real-time event store-and-share across distributed event processing networks is helping drive this work along with real-time BI and Enterprise BAM.
• Are there any use cases for CEP in advertising placements and search engines?
  Answer: None that I recall (but I’m sure some blog readers may correct me… )

Notes:

[1] I didn’t add the obvious corollory that the Business Rule industry could possibly benefit from a similar cross-vendor/analyst/academia society to EPTS - currently the only non-adverserial meetings between rule engine vendors are when the likes of TIBCO, Oracle and Ilog meet at OMG and W3C standards bodies…



Source...