Design optimization of multi-cluster embedded systems for real-time applications
Paul Pop, Petru Eles, Zebo Peng, Viacheslav Izosimov, Magnus Hellring, Olof Bridal · 2004
An increasing number of real-time applications are today implemented using distributed heteroge-neous architectures composed of interconnected networks of processors. The systems are hetero-geneous not only in terms of hardware components, but also in terms of communication protocols and scheduling policies. Each network has its own communication protocol, each processor in the architecture can have its own scheduling policy, and several scheduling policies can share a processor. In this context, the task of designing such systems is becoming increasingly important and difficult at the same time. The success of such new design methods depends on the availabil-ity of analysis and optimization techniques. In this paper, we present analysis and optimization techniques for heterogeneous real-time embedded systems. We address in more detail a particu-lar class of such systems called multi-clusters, composed of several networks interconnected via gateways. We present a schedulability analysis for safety-critical applications distributed on multi-cluster systems and briefly highlight characteristic design optimization problems: the partitioning and mapping of functionality, and the packing of application messages to frames. Optimization heuristics for frame packing aiming at producing a schedulable system are presented. Extensive experiments and a real-life example show the efficiency of the frame-packing approach. 1