Jim Sinur, newly returned as analyst for the Gartner camp, just made an interesting post on rule representations. As TIBCO is attending the OMG and W3c rule standards meetings this week, we’re probably in a good position to critique Jim’s view (and make some, er, suggestions).

Lets start at the top. Jim says business rule experts can choose from a variety of representations:

• “semantics” (by which I think Jim means formal vocabularies rather than OWL-type ontologies and things like Common Logic)
• decision tables, as offered by most BRMS tools include TIBCO BusinessEvents‘ Decision Manager
• “rule hierarchies” which may mean decision flows or decision graphs [*1]
• natural language, which usually means writing rules in any way you want, but could also imply using Ron Ross’ Rulespeak or the OMG SBVR standard, or a constrained language specified by a rule management system for operational (executable process) rules.

Jim then starts talking about rule *execution* rather than business rule representation. I should be clear that “business rules for business people” as defined by BRG et al refer mostly to policy-derived rules, which may be used to guide the development of executable business processes and their associated decisions (such as process gateways or decision services). Typically such policy-type “business rules” are at the OMG MDA computation independent model level. Executable business decisions, on the other hand, typically are at the OMG MDA platform independent model level, which is an IT layer. Such decisions or operational rules may be embedded in process models like OMG BPMN, or (ideally) kept as shareable, declarative production rules per the OMG PRR standard - probably what Jim means when he says “stand-alone rules that can be executed in dynamic sequence where rules are reused in different sequences” (although, Jim, please note that there are precious few rule engines that can forward and backward chain using the same rule definitions). Complex Event Processing is a type of automated process than can use production rules as in PRR.

Jim goes a bit off-track talking about OMG BPDM. Perhaps he was thinking about something else, because BPDM is the Business Process *Definition* Metamodel, not Data Model, and is meant to be a “metaprocess” model [*2].  And it doesn’t have much to do with business rules, other than potentially sharing vocabularies with SBVR and possibly referencing PRR type rules. The latter is unproven at this point in time, but BPDM is interesting simply because of its development potential to provide, in some future version, a shared abstract behavior for both simple and complex event processing. It certainly deserves more support from academic research teams.

In other, related, news from OMG this week:

1. Apparently the BPMN 2.0 notation is to include more sophisticated event handling. We’ll have to wait until we can review the latest draft, but it will be a pleasant surprise if this provides some CEP modeling capability out-of-the-box.
2. PRR made some good progress, and will hopefully move to Finalization Task Force 2 to complete - with thanks to Robert Ong at / and MagicDraw for their metamodeling re-work. One of the Forrester analysts endorsed this PRR work this week, too.
3. There was an interesting panel (OK, it was the *subject matter* that was interesting, not us panel members) on the relationships between rules, processes, and business architectures. And we could have thrown in business events, too. My main points were that “rules” meant both SBVR-type policy rules and PRR-type automated operational rules / decisions, and that vendors need customer support to encourage them to develop standards. And that OMG should not be incubating standards for the sake of it, certainly not without academic contributions.
4. TIBCO was one of 2 vendors providing feedback on Dynamic Business Activity Modeling. This includes iProcess Conductor, goal- and plan-driven business processes, usually teamed with TIBCO BusinessEvents for decisions on plan selection, plan creation, and plan automation - including event-driven replanning. Also TIBCO BusinessEvents itself is used to model high-level event-driven rule-based state models to drive BPM workflows from business events.
5. As a consequence of the above, it looks like there is interest in a standard model for Case Management. Supporters are to gather at the next OMG meeting to discuss what this will entail…
6. Looks like Event Metamodel and Profile and Agent Metamodel and Profile will be issued as RFPs. Any TIBCO customers interested in these, please let us know via email or TIBCommunity!
7. EPTS membership discussions with OMG started.
8. Good to see Rob James (HSBC) of NRL fame at the meeting (another Natural Language for rules for Jim’s readers to consider, and a possible candidate to help automate SBVR type rules).

   UPDATE

9. OMG passed the FTF2 resolution for PRR to get onto its final track, with new additions to the FTF2 team from Inferware, 88Solutions and Unisys.
10. W3C made good progress on some of the RIF PRD (production rule dialect) with many of the same people in PRR. PRD contributors included TIBCO, Ilog, Inferware, RuleML, Fair Isaac and Oracle.

 Notes:

[1]  Interesting to see one of the few vendors to provide decision graphs has just been acquired. Is there anyone else left out there?

[2] I attended some of Conrad Bock (NIST)’s “advanced BPDM tutorial” this week, so am reasonably confident about this!

… has hit the newswires [*1,2,3] along with an IDC comment that TIBCO BusinessEvents is “the undisputed CEP leader, with a market share of 40.2 percent”. TIBCO BusinessEvents version 3 has already been in the hands of customers for a few weeks, and has lots of interesting cool features, such as:

• Distributed decision engine - for real-time event-driven decisions
• Multi-agent blackboard architecture - for complex Complex Event Processing, reliability and failover support, etc
• Workflow-maintained decision management - for rule-driven control of the lifecycle of decisions
• Support for multiple extensible event processing paradigms, including inference rule, continuous queries, and state management.

We’ll expand more on these over the coming weeks, but for now we need to start defining TIBCO BusinessEvents as an event processing platform, not just an “engine”. The platform covers:

• Event processing via rules, queries, custom algorithms, etc
• Distributed persistence of event history via a high-performance data grid
• Reliability and scalability via multiple agents
• IT modelling via concepts, states and rules
• Business modelling via decision tables
• Multiple event channels supported for flexibility.

Links:

[1] Infoworld / PCWorld Business gives some more analysis, and comments on the follow-up step being event analytics.

[2]  ZDNet’s Dana Gardner makes some excellent points on the news and why the CEP space is important. And ends with “… TIBCO’s products are pointing up how now to take the insights of CEP into the realm of near real-time responses and ability to identify and repeat effective patterns of business behaviors. Dare I say, “agility”?”

[3] TechTarget adds another view concentrating on the rule management feature.

So “Event Processing Week” is over. We had 2 days of Gartner Event Processing conference, 2 days of Event Processing Technical Society, and a last day covering EPTS Business and Adrian Paschke’s DoReMePat EU R&D Project submission. Next year will likely see Gartner move their EP event alongside their AADI summit (a good move) in Dec09, and EPTS will possibly convene at DEBS09 in Vanderbilt in Jul09.

Observations?

1. There were a few presentations that made it appear that CEP = continuous query languages. Sorry folks, thats not the case, and hybrid approaches are the future.
2. Roy Schulte made a key comment: we need to make business decisions faster and better. If nothing else, Gartner should be congratulated for supporting EPTS and supporting the industry in this way.
3. Event Processing is not slowing down in its adoption. Expect the high rate of new announcements from the industry to continue. Expect new players to join in from the specialist BI, BAM, BPM, middleware, DB vendors.
4. There are still topics to be explored - event provenance and modeling, for example.
5. A decent Event Processing taxonomy is still needed. Why should anyone think that “EP” was any different from processing events in BPM, for example? Complex “Event Processing” is as good a term as “Complex Event” Processing, and indeed makes it clear that we are solving non-trivial problems. Businesses *know* they have complex event-related problems to solve.

TIBCO will be helping charter the EPTS working groups on reference architectures and standards - its going to be a busy year!

Notes: other articles and blogs covering Gartner and EPTS included:

1. Brenda’s Business Driven Architect covered EPTS days 1 and 2a, 2b, 2c, 2d.
2. Seth Grimes’ Intelligent Enterprise article on Reuters News processing from the Gartner EPS.
3. Doug Henschen, also on Intelligent Enterprise, covers Dashboards, Decisions and Wall Street from Gartner EPS.
4. Opher at IBM covered Gartner and EPTS .
5. Louis at Apama seemed to be amazed that vendors actually spoke to each other! (Its not “war”, Louis, just features, functions and benefits!)
6. Mark at Streambase simply referred thoughts about Gartner to Marc at Citi / Magmasystems. Marc blogged a few thoughts on Gartner, but no doubt was distracted by other stuff going on in his industry.
7. Colin Clark, who used to be a “CEP vendor person” himself, covered Gartner days 1 and 2. Interestingly, Colin blames the lack of attendance at the vendors’ door for not selling themselves as CEP Platforms. But vendor offerings are not (as per normal) discussed in the Gartner event advertising. But if he knows anyone who needs a real, distributed, multi-type CEP Platform he should get in touch!

The last sessions on the last general day for EPTS4 covered… standards! Chris Ferris, who leads IBM’s Standards group, presented a background on why standards are important. His main point was that standards can persist much longer than you intend, using the case of Roman chariots (100BC) leading to rutted roads and standard axles (100-1800AD) and railroad tracks (1850 on) and thence tunnel sizes which determined the maximum size of Shuttle boosters (1970s on) as they travel on the train network.

Girded with high-caffeine Pepsi and Hotel Coffee, the attendees watched a demontration of concensus in the ensuing standards panel, managed by Gartners’ Roy Schulte. TIBCO’s presentation (in the set 7 minutes, more of a projectile vomit of pointers on the topic), covered:

• standards are not vendor inventions, but open vendor collaborations usually within an existing standards body like OASIS, W3C or OMG.
•  Possible standards include messages / middleware (like JMS), Event Processing Elements / pattern languages, Event Processing Agents / deployment APIs, user controls like rule management languages, and visualization types.
• Middleware standards are probably not the domain of EPTS - event delivery is a different domain from event processing.
• Standards don’t just “happen” - they cost time and effort, and need to be specified by customers (who ultimately pay for their development).
• As CEP is not a “complete change” from conventional IT, existing standards and efforts are relevant:
  • UML Class and UML State for modeling
  • OMG PRR and W3C RIF are nearing completion, and can be easily extended for event-driven rules
  • The new OMG EMP for event models is about to hit RFP stage, and is of direct interest to EP vendors
  • Ditto the new OMG AMP for agent models, of interest to distributable EP Network vendors
  • For Continuous Queries, it is time for the ESP vendors to get their act together and define a baseline. Before someone else does without our input.
  • Orchestration diagrams for event processing elements is quite a common feature, yet could be standardised by a version of BPMN (either extending BPMN or doing an EP version).
• Standards bodies are already creating interest in CEP standards: at OMG next week for example, not only is PRR, EMP and AMP being discussed, but also a Business Modeling and Integration discussion on Dynamic Business Process Modeling (covering EP and BPM), and a Business Architecture discussion on the intersection of rules, processes, etc.

To end up, Peter Niblett (IBM) gave an update on WS-Eventing (W3C, Microsoft) and WS-Notification (OASIS, everyone else) and the defunct end of the attempt to join them (WS-EventingNotification). It seems the general opinion is that the WS-* stack has been overloaded, and is not ever going to be associated with EDA. Although IBM supports it, no-one else does, nor will they unless customers think it is useful and demand it.

The next sessions from EPTS4 covered some academic research projects.

• Prof Adrian Paschke (newly promoted to Berlin Uni) covered RuleML, Reaction RuleML, and the related standards of W3C RIF and PRR, then Alex from Betfair covered the Prova project which used state machines mapping to a PROLOG-type language. This was an excellent introduction to the rule standards space, albeit with the usual complaint that Adrian (and the other RuleML advocates) really should be clearer about the role of RuleML as a common research platform, not a commercial standard competing with W3C etc. [Disclosure - TIBCO participates in both PRR and RIF standards].
• IBM US R&D covered their designed-for-scalability stream processor, System S. I would guess the existing stream processing vendors would be pretty nonplussed by this…
• IBM Israel R&D covered the history and directions for their Event-Condition-Action Amit CEP research engine - with an interesting innovation of a real-time event-handling IDE (presumably combining monitoring with in-situ pattern specifying).
• Michael Olson of CalTech described some more about their social aspects research for EP, such as semantic web and mashups using event processing, and the personal event broker.
• Nenad Stojanovic of FZI went through his CEP for Agile BPM - an area we are already seeing in the commercial world, but worthy of more research - such as BPMN for CEP, etc.
• Pedro Bizarro (Uni of Coimbra) covered his benchmarking project, which probably has more potential than the STAC commercial effort to look at more event pattern benchmarks, get customer project data contributions, etc.
• Suzan Urban (Texas Tech Uni) described their StreamCEDL work.

One thing the academic projects should seriously consider is exploiting / extending existing commercial software frameworks. Imagine how much easier it would be for a benchmarking effort to build a generalized framework if using multiple existing CEP tools, or the StreamCEDL work if it used an existing distributed CEP platform. TIBCO’s CEP platform for example is available via the University Program, and interesting projects could be discussed with customers and experts via TIBCommunity. Just a thought.