This year, Daniel starts with a disclaimer that we are not allowed to talk about ILOG any more… It is IBM… I will miss the name though.
Daniel covers the many technologies that exist in Business Applications for decision management. Namely, this is about the characteristics and interplay of BRMS, BPM and CEP. I am delighted to have him cover this topic. As I commented in an earlier post, prompted by a brilliant presentation made by Gartner analyst Jim Sinur at BPM Summit, we need more clarity in this space.
Starting with Processes
Quoting Sandy Kemsley, Daniel maps BPM (Business Process Management), DCM (data Center Management), ECM (Enterprise Content Management) and ACM (Adaptive Case Management) in a data /process chart. The interesting piece here is the evolution in terms of capabilities ranging from fully structured to fully adaptive.
Policy Changes: Medium
Model Changes: Hard
Requires:
- Retrain humans
- Update diagrams
- Update screens
- Migrate process instances
- Update test data
Moving on to Decisions
Daniel proposes a definition, that he is not too happy with, allegedly, but at least it is a definition: “Process of selecting from several choices, products or ideas, and taking action”
State is much less important as a decision is often a point in time activity based on available data.
I am not so sure I agree with Daniel’s statement that rules are not very reusable though. By definition, decision services are reusable of course — the fabled Universal Decision Engine. Smaller chunks of decisioning logic have also been reused with much success with a smart rule architecture.
Policy Changes: Easy
Model Changes: Medium
Requires:
- Update business rules
- Update test data
And then Event Processing
Complex Event Processing definition has been pulled from Wikipedia.
CEP applications tend to be stateful within a short time-window. The more events rushing into the system, the shorter the time-window.
The implications on architecture are obviously significant: State persistence, Data enrichment and/or transformation, continuous query engines, etc.
Rules are tricky to reuse because they have tricky pre-conditions.
Policy Changes: Medium
Model Changes: Medium
Requires:
- Update filters/actions/queries
- Update test data
And finally my son’s favorite part: playing lego!
Those technologies can be combined in many different ways. The most common patterns end up being those two:
- They can used independently of course — separate systems manually integrated => requires skilled IT organization to manage the complexity
- Decision + Events (leave BPM outside) — attractive because those 2 technologies are all about rules (stateless and stateful)
Questions?
Paul Haley asks again on elaboration re: failed efforts. Real-time pattern detection? We’ll see if it is actually a trend…
Learn more about Decision management and Sparkling Logic’s SMARTS™ Data-Powered Decision Manager