Ontological specification of performance models for multi-agent systems

Large-scale multi-agent systems (MAS) utilize performance models for their optimization. These models could be queueing network models such as Jackson or BCMP networks that can capture the flow of tasks through the network of agents and estimate quantities such as response time and queue lengths. The selection of the model and subsequent analysis of the system is often done offline. We propose a generic ontological framework to capture the semantics of various performance models to facilitate online analysis of MAS with minimal human interaction. Through the framework, we illustrate the specification of queueing models used for MAS performance optimization.