The increasing complexity of distributed computer systems requires new control mechanisms. The behavior of future systems should be defined by high-level goals, with the system itself being responsible to maintain them. This paper proposes to use the constraint satisfaction problem (CSP) paradigm to realize such self-configuring systems. This allows to specify the desired system behavior as constraints and generic domain independent algorithms can be used to enforce these constraints. We present a novel algorithm called System-Driven Search (SDS) for maintaining constraints in highly dynamic distributed environments, like wireless sensor networks. It is not susceptible to message loss and piggybacking may be used for information dissemination instead of sending explicit messages. These features make SDS suitable especially for decision making tasks like self-configuration in battery-operated wireless sensor networks. Partitioning into coordinating cliques and channel allocation, two typical problems in that area, are used to evaluate the proposed algorithm.
展开▼