Model-Based Reconfiguration Planning for a Distributed On-board Computer

Abstract

The ScOSA project (Scalable On-board Computing for Space Avionics) of the German Aerospace Center aims at combining radiation hardened space hardware together with unreliable, but high performance COTS (commercial off-the-shelf) components as the processing nodes in a heterogeneous on-board network in order to provide future space missions with the necessary processing capabilities. However, such a system needs to cope with node failures. Our approach is to use a static reconfiguration graph that controls how software tasks are mapped to the processing nodes, and how this mapping should change in response to possible node failures. In this paper we present a model-based approach and a tool for automatic generation of reconfiguration graphs. Based on the software and hardware models, we traverse the graph of all possible failure situations. For every node of this graph we solve a combinatorial optimization problem of mapping tasks to processing nodes either with an SMT solver or using a genetic algorithm. The resulting reconfiguration graph can then be translated into the configuration files that are deployed on the target system, eliminating the need for tedious and error-prone manual configuration design.