An Incremental Algorithm for High-Performance Runtime Model Consistency

We present a novel technique for applying two-level runtime models to distributed systems. Our approach uses graph rewriting rules to transform a high-level source model into one of many possible target models. When either model is changed at runtime, the transformation is incrementally updated. We describe the theory underlying our approach, and show restrictions sufficient for a simple and efficient implementation. We demonstrate this implementation in Fiia.Net , our model-based toolkit for developing adaptive groupware. Developers using Fiia.Net control components and connections through a high-level conceptual runtime model. Meanwhile, the toolkit transparently maintains the underlying distributed system, and propagates failures back into the conceptual model. This approach provides high stability, and performance that is sufficiently fast for interactive applications.