Brenda Michelson announced that the OMG EPCS webinars are up on the web: register here for access links. Presentations included:
- Gartner’s Roy Schulte on “The Time for Event Processing is Now” (a pun on the fact that the time for event processing is always now, otherwise we are doing historic (data) processing…)
- Roy introduced the concept of being “event-driven”, and CEP leading to “situation awareness”. Roy’s version of what we call the “event – decision – action” model is the “event – compute+analyse+decide – response” model – of course these are equivalent with some fuzziness between the event and the decision…
- Conventional (transactional system) reports and BI are not event driven, and suffer from too much “business latency” – problems are detected after they occur rather than as they occur, leading to fewer opportunities for intervention.
- Roy’s CEP patterns included:
- the “operations or process monitor” (a.k.a situation awareness in the business operations and processes)
- the “supply chain management monitoring” (a subclass of the above)
- the “enterprise nervous systems” such as monitoring airline operations or large railroads
- The EP Market viewed “continuous intelligence” applications often using custom CEP apps is >$10B in 2009, with the BAM and CEP tool market being $189M in 2009 with likely growth of 30-40% pa.
- Chris Bird at SABRE showed an “Event Distribution Architecture” case study, referencing the book and analyst methodology “Lost in Translation” that is based on policies, events and content.
- Events are not necessarily about “real time” behavior, but responding as required. The “content” is the information related to the event, and issues include “how to avoid gumming up the works”… so the “enterprise pattern” is to define events, interested parties, information and security models… leading to a “policy event content” framework.
- On events, Chris assumes separate content and control channel to coordinate (transactional?) event processing, with a “policy engine” monitoring the control channel.
- Chris’ explanation of his view on event processing seemed definitely about coordinating “event transactions”, talking about “message failures” and the suchlike.
- Content structures could be XML, legacy formats, or some canonical form – Chris proposed sending both old formats and new canonical forms, or possibly transforming the event in a “staging processor” and republishing for other processing agents, giving much flexibility.
- Chris’ key points were to concentrate on event distribution, not event processing, and loose coupling was better than tight schemas…
- Opher Etzion of IBM Research presented on the next 7 years in event processing
- Trend 1 was from narrow to wide focus applications, beyond the current domains to robotics, home automation etc… although to be fair most of the embedded systems guys already do lots of event-driven state modelling in their systems.
- Trend 2 is from monolithic to diversified systems.
- Trend 3 is from proprietary to standardised, for example in languages allowing for interoperability.
- Trend 4 is programmer-centric to semi-technical centric – hence modelling, high level languages like event pattern languages, and so on.
- Trend 5 is from standalone packaged applications to embedded components.
- Trend 6 is from reactive to proactive.
- Directions are that multiple types of platform are moving to a virtual event processing platform, a movement to tailor-made optimisations within event processing,and event processing software engineering.
Overall some interesting points made in these presentations, and the event was an interesting alternative to the more academic DEBS and Dagstuhls this year.
[Disclaimer: TIBCO presented too. I’ll leave the commentary on that to other folks!]
[Update: removed one review due to author objections!]
