Cognitive Grid Part 3:
The Cognitive Grid Methodology.
The methodology of the Cognitive Grid is founded in classical OO design, but draws critical extensions from Agent-Oriented design, Semantic modeling and Distributed, Fault Tolerant Networking. The methodology is principally focused on defining the actors, processes, relationships and communities towards identifying the components and configuration to realize a dynamic intelligent system. While this sounds complex, it is actually a more natural way to map a system the OO since your principal elements are closer to those people, organizations, business processes and agreements found in the physical world. Effective automation should be oriented toward natural representations and replacing physical components with computational ones. While drastically simplified for this overview, the basic steps to the methodology for design are:

1. Enumerate and define each of the component actors, splitting actors until they have one primary functional responsibility of medium complexity.
2. Enumerate and define the primary and seconday processes, representing the process flows as state diagrams and utilizing sequence diagrams for information and tasking flow.
3. Enumerate the permanent and transient relationships between the actors or logical communities, including key characteristics like command, support, negotiated, ad-hoc and discovered.
4. Cluster the actors into logical communities, with every actor in at least one community.
5. Decompose the processes into components, where each component can be a encapsulated, reenterant unit of behavior.
6. Using the understanding of 1-6 above, design a hybrid object-semantic model with annotations for database persistence, source data, properties and types.
7. Develop the specifications of each behvior component, including the services utilized, pre/post conditions and model components utilized
8. Map the components back to the actors, and actors to communities, nodes, data sources and devices, this becomes your deployment template.
9. Map the nodes to machines, this becomes your deployment configuration.
10. Build, Integrate and Test as components, actors, communities, nodes and ultimately the full system.
11. Iterate over 1-10 as necessary.
12. Deploy the system.
13. Periodically revisit 1-11 to refine existing or add new functionality, updating the deployment when appropriate.

The real key is thinking about the system a set of actors, each having behaviors, working together to perform business processes for, and in concert with, the operators. Properly leveraging the semantics of the model, the reasoning components of the fusion engine and situational engine, as well as designing effective user automation screens are part of the art of designing effective intelligent systems, but can be learned with a little practice. Remember, the mechanics of the methodology need to be performed under the guide of the philosophy - be ever mindful of the five core principals.