Monitoring Mechanism Design for a Distributed Reconfigurable Onboard Computer

Abstract

The Onboard Computer - Next Generation (OBC-NG) project is initiated by the German Aerospace Center (DLR) to develop a distributed and reconfigurable onboard computer of Commercial off-the-shelf (COTS) components. Its goal is to utilize the higher computational power of COTS components and to maintain high system reliability, which is required for spacecraft. However, COTS have a lower robustness than the traditional space-qualified components and therefore a higher probability of failures. Thus, a monitoring system is indispensable. This thesis focuses on the design of a monitoring system, to evaluate the health status of the distributed components. It shall detect failures before any corrective action, e.g. migrating the tasks to the other functioning components, can be triggered. Different monitoring techniques have different trade-offs, in terms of monitoring efficiency and monitoring overhead. Initially, various monitoring concepts are investigated, in order to qualitatively analyze their trade-offs and designs. Three monitoring mechanisms, PULL, PUSH and PUSH-PULL, are selected and modeled. Afterwards, the models are simulated on the discrete event simulator OMNeT++ and tested under different environment as well as monitoring mechanism settings. The simulation results, which represent a quantitative investigation of the designed monitoring mechanisms, are used to evaluate and compare them. The derived verdicts are used to identify the most suitable mechanism and settings for the OBC-NG system. The results show that PUSH mechanism is more suitable for OBC-NG system than the currently implemented PULL mechanism.