Graph Trace Analysis Approach to Optimizing Power and Heat Flow for Clustered Computing; An Example of Model Based System of Systems Design and Deployment

During system of system (SoS) design, many decisions from different engineering disciplines are made and documented. Today, designers increasingly use modeling. For example, they address the issues involved with total cost of ownership by better understanding the interactions of parts and systems. For a large SoS, the labor intense modeling has one potential alternative approach involving the concepts of graph trace analysis (GTA) for distributed processing of integrated system models (ISM). GTA features a wide range of algorithms that flexibly attach to models that stay in their \´engineering domains\´ in the ISM. A subsystem model for a single domain can be built directly from the engineering design data and then simulated or analyzed. A model using millions of simple objects can be built automatically, including looped and radial systems. Physical dependency linkages between engineering domains can be built directly from reference designators and parts data. Design and deployment of a compute cluster is illustrated by composing an "integrated system model" (ISM) from different engineering domain models, where models from different domains are linked together with dependency iterators from GTA. During early design phases, the ISM of a compute cluster might include a short list of engineering design domains. Later, during deployment, sensors and actuators are included in the ISM and GTA algorithms. A need for open, standard parts data is discussed for design tools used in engineering along with data exchange formats that allow product design data to be reused by other design teams.