BPAF: Possible base event data model for SAMM

[heiko.braun]

This might seem out of context for most of you, however in a recent discussion with Gary we were talking about a base model for activity monitoring in a larger context. Here's something I stumbled upon which we might lean on:

2 Minute Intro:
http://www.bpm-research.com/2009/02/22/the-business-process-analytics-format-bpaf/

30 Minute Intro:
http://www.bpm-research.com/wp-content/uploads/2009/02/2009-02-20-wfmc-tc-1015-business-process-analytics-format-r1.pdf

[camunda]

Hi Heiko.

Just a quick feedback: Interessting!
In the project of 1&1 where I worked in we developed some events pretty similar to this (you can read about it in the next JavaMagazin in Germany). Which maybe proves it is not a theory-only-thinkg :-)

Cheers
Bernd

[heiko.braun]

Yes, I consider "process" and "service" as exchangeable terms in this context. So on the one hand, a common event format could be used to store, query and visualize history information, and on the other an activity monitoring tool like SAM could benefit from a common base model across services (as in ESB) and processes (as in BPM).

What are your experiences with it? Especially with representation of custom payload (bpaf:dataElement) attached to a base event?

[camunda]

We attached not too much information, just the most important business keys (customer id, order id or the like).

This makes it for example possible to see all running processes (over different instances) of a customer centrally.

[objectiser]

Hi Heiko

Does look interesting, although I think we could have a base event at a higher level of abstraction, just including attributes EventID, ServerID (optional) and Timestamp, as we discussed.

Therefore any activity could be reported based on the base event, and any 'process/service' related event could be based on this model.

Regards
Gary

[heiko.braun]

We attached not too much information, just the most important business keys (customer id, order id or the like).

I was wondering if that is sufficient. Obviously just allowing name/value pairs simplifies both storage of that data and querying it later on. We'd also been thinking that maybe just storing a reference id to access the actual payload externally (i.e. Order database) might already be covering a lot of the use cases.

What do you think?

[camunda]

Hmm, key / value pairs may be an option, but the database schema gets already much more complicated. In this project we knew the 4 important keys for the company, so we just added them and have everything in one table.

But this is indeed not the generic approach, which I guess you are targeting.