Enterprise Architecture: From Incite comes Insight...: Outstanding questions regarding BPEL and ESB

Monday, January 16, 2006

Outstanding questions regarding BPEL and ESB

My son, absolutely loves Bob the Builder and tells me that it takes teamwork to get the job done. I previously asked questions on Rules Engines, Voice over IP and even Enterprise Content Management (ECM) and hope that an interesting discussion on BPEL and ESB may emerge...

ESBs such as ServiceMix support two features which are very important to enterprises that I haven't seen industry analysts talk much about. The first is support for rules-based routing. ServiceMix incorporates the Drools rules engine. I suspect there are many situations in which static routing runs out of limitations and hope that the industry will start blogging on the notion of dynamic routing. The second thought is related to supporting streaming APIs such as StaX as DOM based approaches for especially large XML payloads tends to slow down performance. Would love to know if ServiceMix is king of the hill in this regard and whether commercial ESBs will be adopting soon?

Some folks within the community use the words integration and interoperability somewhat interchangably. I am of the belief that they are different but couldn't find a crisp definition that separates out their distinctions

Curious when one should use a BPM with an ESB. It seems to me that if you have an ESB with a native service model, then the concept of BPM is already built-in. How come BPM vendors aren't talking about ESB?

One of the drivers of SOA is decentralized execution, doesn't adopting a BPM engine and all of the licensing constraints in may supply that force centralization result in an inefficient architecture?

The notion of Business Activity Monitoring (BAM) seems to be more aligned with BPM. How come it can't be easily realized using ESB constructs?

Which BPM vendors are currently working on certifying themselves to run on 384 CPU Appliances such as Azul Systems.

In theory all BPM engines can sit on top of all ESBs. In reality, this simply doesn't work. Is anyone working on an interoperability consortium to solve for this problem?

Some of the BPM vendors suggest that they can assist with Six Sigma efforts. Any of them planning on publishing a publicly available reference architecture or at least willing to blog about it so we can understand how this may be implemented?

Shouldn't Feugo be showing up in more magic quadrants? In fact, I would think they would be listed ahead of many of the current players because they have useful capability that others dont. They have incorporated Neural Nets into their BPM Product. A neural network works through a decision activity capability that lets users define a set of variables that can be analyzed
for process improvement.

If Neural Nets allow for cognitive modeling, wouldn't it be better to not listen to hype regarding BPM modeling as it is highly limiting? I wonder what Fuego could do to help us Enterprise Architects in corporate America think about better ways to model?

Many enterprises hear of the wonderful story told by vendors regarding BPM but never hear of the limitations of BPM. For example, BPM doesn't really help capture semantics, bridging of security models or other hard problems and in many cases simply becomes a mechanism for chaining together sub-optimally designed services. Are there any BPM CTO's or industry
analysts in this space willing to tell us of other BPM limitations?

John Evdemon has always pointed out that enterprises are thinking about BPEL incorrectly. Instead of thinking about it as a programming language for business process they should consider that BPEL is about interoperable abstract processes - not portable executable processes. The rationale for this thinking is the specification misses many fundamental concepts.

Awhile back, I read that there were published well-known defined workflow patterns and that the only vendor that implmented them all is jBPM. Does this story still hold true today? How come the analysts aren't pointing out patterns-based approaches to BPM? If an enterprise wanted to adopt a BPM engine, wouldn't it be better served by ignoring quadrants and instead choosing the one that implemented the most documented patterns?

I periodically interact with folks from the Object Management Group (OMG). Awhile back they were pushing the notion of Model Driven Architecture. Maybe someone can tell me why I can't have MDA Driven BPEL? While they are at it, how about telling us why I wouldn't want to use MDA as an approach vs. BPMN?

There are a ton of WS-* specifications, many of which touch upon BPM. How should I think about WS-BrokeredNotification, WS-BusinessActivity, WS-Choreography and WS-Coordination with BPM? Any BPM vendors implementing these specs or at least have some thinking on them?

How should I think about integrating Liferay Enterprise Portal (100% open source for life) with a closed source proprietary BPM engine?

Has any industry analyst ever made a comparison of the BPEL that is built into ServiceMix and ActiveBPEL? Both are free in terms of acquisition price for an enterprise. I wonder when analysts will start making these types of recommendations to us folks in corporate America?

How should I think about BPM and Business Rules. Do they really belong in the same engine? Someone should provide me with a publicly available reference architecture so I make an intelligent choice here. Speaking of this, any thoughts on a Service-Oriented approach to Business Rules?