Canonical Model, Canonical Schema, and Event Driven SOA

re: Canonical Model, Canonical Schema, and Event Driven SOA

Ashi — Wed, 13 Jun 2007 11:44:27 GMT

It is an effiecent & Agile way to break the Business Model/Architecture.

I agree with you that, as our SOA implementation reaches some maturity we do tend to think in these lines. But keep these in mind before hand, will save us a lot of Time and Confusion.

Thanks for relating these concepts..

re: Canonical Model, Canonical Schema, and Event Driven SOA

Jack van Hoof — Wed, 13 Jun 2007 18:54:54 GMT

GREAT POST!!!

Now you definitely are my friend, Nick...

From this viewpoint EDA might be seen as of a higher architectural magnitude then SOA as the eventing pattern puts constraints on the services and not vice versa: http://soa-eda.blogspot.com/2007/06/magical-of-soa-and-eda.html

re: Canonical Model, Canonical Schema, and Event Driven SOA

Kjell-Sverre Jerijærvi — Thu, 14 Jun 2007 22:38:12 GMT

I still don't believe in "one true schema", but we're one the same page wrt EDA and semantic covenants:

http://kjellsj.blogspot.com/2007/06/semantics-covenant-service-is-always.html

re: Canonical Model, Canonical Schema, and Event Driven SOA

Jack van Hoof — Mon, 18 Jun 2007 17:05:31 GMT

I don't think this is about "one true schema", but about a mechanism to map different schema's to be able to pass semantics across different environments:

http://soa-eda.blogspot.com/2007/04/how-to-mediate-semantics-in-eda.html

What else would you suggest to accomplish this, Kjell-Sverre?

re: Canonical Model, Canonical Schema, and Event Driven SOA

NickMalik — Mon, 18 Jun 2007 21:15:05 GMT

@Jack

I could not have said it better myself.

@Kjell-Sverre

The goal is to figure out how to communicate. Think of this like the diplomatic community. In a country, a lot goes on that the diplomatic community is not really worried about. However, when we want to talk between ourselves (to create an international treaty, for example), we need to have a common language to negotiate and sign the treaty in. That common language (not the content of the treaty) is the stuff defined by the Enterprise Canonical Data Model.

re: Canonical Model, Canonical Schema, and Event Driven SOA

Rob Eamon — Fri, 22 Jun 2007 20:50:44 GMT

+1. Like Jack, I really like how you've laid out the concepts. Very nice. App-independent messages are a key decoupling mechanism.

"Creating this bird is not easy. It requires a lot of work and executive buy-in."

Indeed. The difficulty in creating ECDM and ECMS cannot be overstated, IMO. This can be really hard--especially when the diplomats in the community aren't all that interested in participating in the exercise--"just send me the data I need". "Didn't we just do this for data warehousing?"

Lastly, you touched on this a little, but this exercise should not lose sight of the business processes. Only in the context of the processes do the events make sense. Integration, IMO, is best served by a "process first" approach. Events and services come after.

-Rob

re: Canonical Model, Canonical Schema, and Event Driven SOA

NickMalik — Fri, 22 Jun 2007 21:43:12 GMT

@Rob,

I agree with a process first approach and I think that's one reason why the business event ontology is so important.

That said, processes have a heirarchy. We speak of Level 1 processes like Marketing, Sales, Fulfillment, etc. Level 2 processes would be under one of those top level ones. For example, under Marketing would be things Create Market Strategy, Segment Market, Build Programs, Execute Programs, Capture Response.

That is the level where the business event ontology really hits home. This is because these Level 2 processes are the domains of large systems. You will tend to find a system that spans the process, largely from end to end. The business likes looking at data within these buckets, and cares a lot less about the individual data elements flowing between them.

So I agree, we start with process. On the other hand, I caution teams not to go all the way down to level 4 and level 5 before starting on Integration and Services. While there are likely to be services developed to support level 4 processes, they will be built in the context of a single system and don't need to be architected "from the center." They need to be architected in the project that builds or maintains the systems that serve those needs.

Process First, but stop before you go too deep.

re: Canonical Model, Canonical Schema, and Event Driven SOA

Kjell-Sverre Jerijærvi — Mon, 02 Jul 2007 09:21:59 GMT

It was the part about the CDM being like a shared relational database that lead me to think about "one true schema" and other common data model approaches. Jack's post makes the distinction between the common data format and the message metadata CDM, and I recommend reading his post first :)

What are the requirements for enterprise event security?

Inside Architecture — Wed, 18 Jul 2007 16:31:25 GMT

In a SOA, to be effective, we need to share both data and events. Events, as I have discussed before

What are the requirements for enterprise event security?

Noticias externas — Wed, 18 Jul 2007 17:27:11 GMT

In a SOA, to be effective, we need to share both data and events. Events, as I have discussed before

re: Canonical Model, Canonical Schema, and Event Driven SOA

Anil Datt — Fri, 27 Jul 2007 21:09:54 GMT

In the practical world the process needs to be put in place may be, by the SOA governance to make all this work effectively. I had been on consulting projects where the enterprise had the Cannonical Model (not the database) in place. The process of subsetting was so cumbersome that we ended up creating our own "Rider" data model and schemas. Probably there could be some kind of tool SOA platform vendors will start providing in the future to ease subsetting,transforming and versioning from the Enterprise repository?

re: Canonical Model, Canonical Schema, and Event Driven SOA

NickMalik — Sat, 28 Jul 2007 06:34:46 GMT

@Anil,

I'd love to hear more. Can you send me an e-mail directly? I'd like to know what, in the enterprise you worked in, made it so difficult to create a subset message from the Canonical Model?

--- N