Middleware specialization for memory-constrained networked embedded systems

General purpose middleware has been shown to be effective off-the-shelf, in meeting diverse functional requirements for a wide range of distributed systems. However middleware customization is necessary for many networked embedded systems because of the resource constraints in the networked nodes. In this paper we demonstrate that reduced middleware footprint can be achieved while maintaining real-time properties of applications running on such systems. We also give evidence that empirical measurement using a representative application is crucial to guide (1) selection of feature subsets from general purpose middleware and (2) trade-offs among different dimensions of design metrics including real-time, footprint, and portability.