The second day of the EPTS meeting in Trento covered some diverse topics such as:

• Network and Systems Management: event processing aspects (including views from Tivoli and CA) - this is already recognised in TIBCO with CEP-based products like ActiveMatrix Service Performance Manager providing much of the same role.
• CEP Research areas had some fascinating topics, including:
  • The role of CEP in Advanced Robotics (Bielefeld University)
  • IBM’s Smarter Planet initiative: basically the need to integrate event processing with business decisions (which may sound familiar) and advanced analytics (also familiar). From TIBCO’s perspective though the main area requiring research here is event-based analytics - determining what analytic algorithms (beyond rules) are adaptable for (or need redeveloping for) the event-based, dynamic, IT world.
  • Some interesting event pattern benchmarking (from FZI) using their own research algorithms as well as open source tools, with some interesting results … possibly this will help restart the EPTS Benchmarking initiative, and for sure the vendors would like to try out the same patterns.
  • Intelligent Event Processing (Toronto University) showed how Markov models could be used in network intrusion detection.

One of the main purposes of the EPTS Symposium was to report the progress of the various EPTS Working Groups. These included:

• Use Cases: EPTS is increasing the range of use cases, and considering classifications to cover in-production and theoretical cases - presenteded use cases included enterprise IT management and using social media stream analysis for marketing support, respectively.  The chair’s grand challenges included exploitng the vast knowledge of the “Management of Data” domain (which some could view as avoiding “re-inventing the wheel”, others as following the same old practices that lead to problematic IT performance), as well as EPTS communications (such as public blog critiques instead of internal communications - no idea what that is about) especially between working groups.
• Reference Architecture: vendors including TIBCO, IBM, Oracle and Streambase presented as well as the host University of Trento. The participant architectures provoked some good Q&A and ideas (for the WG and future blogging!). We stole exploited the Use Case WG example on Healthcare to demonstrate the applicability of the architectures, but there is still work to be done on combining the ideas into a useful form.
• Glossary: the chairs of the Glossary WG were not in Trento but joined by phone. There were some revisions to the glossary under discussion, and it seems the discussion is finally exploiting the EPTS wiki (hoorah for Web 1.5-based collaboration!).
• Languages: Opher gave an update on the work of the Languages WG, which seems to be at a crossroads: metalanguage or EP language classifications?  The WG presentation ended with a long discussion on standards - need, role, level etc. (My 2c contribution was that EPTS needs to avoid developing standalone new standards and should concentrate on supporting updates to existing standards for event handling, both at the domain (e.g. supply chain) and IT (e.g. W3C RIF) levels. And maybe push the Java Community Process (like being done for constraint programming) to compete with MSSI…

There is an EU Project proposal to look at Event-driven Business Process Management, and that drove the title and interest in this panel (following neatly on from the edBPM09 workshop a few weeks back) chaired by CITT’s Rainer von Ammon.

The brief for the panel was to cover particular aspects of  “edBPM” such as:

• What is meant by edBPM? Compared with, for example, the “edBPM  Reference Model” presented at the 1st EPTS symposium in 2006?
  • TIBCO of course talks about “edBPM” as either a part of “BPM+” (from a BPM perspective) or as automated rule- and event-driven processes (from a CEP perspective). Either way, customers regularly combine complex event processing and (orchestrated, BPMN-based) workflow as well as (orchestrated, XML processing) SOA.
  • The main issues with the supplied edBPM model was that it simply combined event processing with BPEL processes. TIBCO’s edBPM customers invariably never use BPEL (why should they?), and indeed some don’t use BPMN for process models (defining models in terms of states and rules). So the main suggestion here is to go up a level in abstraction:
    • Multiple engines (that could include BPEL if you were so hindered inclined) as well as event processing algorithms or even analytics.
    • An event server (or bus).
    • Some kind of generic state store (for process states, persisted events, etc).
    • More generic models (for business control) and dashboards (cockpit or otherwise).
• Who will be the first “market mover” in exploiting the term “edBPM”?
  • This might have been disappointing to the edBPM pundits, but the vendors basically agreed that this term had no formal “legs” yet: no one had a marketing campaign around event driven BPM, no one had tried to persuade an analyst to take up (or define) the term, and the large BPM vendor and consultant community would not want their “BPM” mindshare asset diluted.
• Do we need new/enhanced standards for edBPM? And what is the challenge to insert/combine Complex Events in BPM? Versus say an example model for non-deterministic approaches like Smart Fraud Management in Banking?
  • The new BPMN2 standard has started the process of adding interesting event extensions to BPMN, while there are existing standards for other models (PRR for production rules, UML State for entity lifecycles, BMM for motivations, PMML for analytics, the proposed DMN for decision models, and so forth).
  • The Fraud example shows how some existing fraud products might work, but again is too specific - what if other event pattern detection and event pattern discovery techniques are desired?
• Related to the above point was the proposal for new standards at the domain level for edBPM…
  • “The good thing about standards is that there are so many to choose from” comes to mind here: effort would be better spent event-enabling the existing standards (what are the relevant loads and what combinations of the domain data models are relevant as payloads?)…
• Although the panel did not progress to the next question, it is interesting nontheless: what is required to set up edBPM projects and/or what aspects of edBPM need to be researched further?
  • Probably there is some methodology work to be done on when to use processes, rules/queries or states and how to combine them…
  • Areas of outstanding research in edBPM are probably joint semantics (a.k.a. the promise of BPDM) across different process types (including CEP), CEP-enabled BPMN, and mergers of event operations with (some aspect of the voluminous domain of) the SOA service standards.

The original CITT proposal can be found at http://www.citt-online.com/downloads/EDBPM-IP-proposal.ppt .

So this week we see the 5th Event Processing Technical Socisty Symposium in Trento, Italy. Trento is a pretty, ancient and university town nestling in the “foothills” of the Dolomites, and is a nice location for EPTS to meet, except that many EPTS members have not made the jouney from the US (Opher reports 45 attendees). Which is a shame, given the agenda.

The keynote covered a unified business modeling approach derived from the OMG Business Motivation Model and mapping from goals and their subsequent intentions via metrics to processes, rules and events. They use constraints (defined using UML OCL) to specify goal constraints, which presumably avoids the need to do any SBVR-type business language translations.

Probably the nearest we have to this in TIBCO today is the Advanced Fulfillment Framework (TIBCO AFF) services offering, in which:

1. Goals (representing desired complex events), such as “car delivered”
   [TIBCO AFF: represented by Plans]
   map to…
2. Processes with activities subjected to goal-derived constraints, such as “delivery time must be less than 7 days from start of the process”
   [TIBCO AFF: plans drive processes selected by or created by rules or manually, and subject to change events and goal modifications ].

John emphasized that the business models had an instance level (instanciation of the model for particular goals and processes etc); of course in a CEP environment one would extend this to monitor resources across processes and thereby also handle resource constraints and SLA policies (e.g. by checking resources before new goal instances are accepted).

One interesting comment was that, out of the box, BMM covered too many primitives to be effective. So the Business Model used just a subset of BMM, the rationale being that BMM’s goal was to support governance whereas this model’s focus was on control (for which only a subset of BMM was necessary). I’m not sure the BMM team will agree entirely with this, but then again, simplicity is good.

John’s final comment was that Event Processing needs to be both top-down and bottom-up (i.e. both goal and event driven) as well as adaptive (both goals and events can change instances of processes).

[Disclosure: TIBCO is a member of the Revision Task Force for OMG BMM...]