Research Of Fault-Tolerant Technology On Masterslave Interconnection Bus Of Satellite-based Parallel System

As the rapid growth of data to be processed in satellite-based systems, the satellite-based parallel systems become the research focus. Because the data bus interconnecting each child processing system is very vulnerable to various factors in space environment, fault tolerant design for satellite-based parallel systems is necessary.And because of the asymmetry of the nodes in masterslave bus, the fault-tolerant issues to be consider ed are more complex.So, we specialized in the fault tolerant of this type of bus.Previous researches in fault tolerance of masterslave bus focused on the increasing reliability of the link or setting up the backups of the nodes etc. Due to diversity of the buses, research for the way of fault tolerance about swithing master and slave nodes is not mature. In this paper, according to the characteristics of the masterslave bus, the model of fault tolerance about masterslave bus is proposed. In the model, we focuse on discovery, restore and isolation of the fault for the way of swithing master and slave.Under the constrants of no external hardware arbitration and no internal communication,the election strategy based on differential threshold and strategy based on pre-assignation are proposed. In the strategy about differential threshold, the range of differential threshold for each node entering into master mode without conflict was derived. And in the strategy about pre-assignation, using the ideas of dynamic optimization makes node can dynamically perceive reliability of all the nodes, and provides the basis for master election. This mode can guide the design of the fault-tolerant design for masterslave interconnection bus.Applying the fault-tolerant model and according to the fault tolerance and communication requirements for satellite-based parallel system, two kinds of bus fault tolerant communication systems were designed. Through the establishment of polling mechanism to realize RT node's active communication ability, the problem about the traditional RT node who can not initiate communication was solved. The interconnection structure of PCI Express bus was designed and solved the fault-tolerant communication problems based on shared memory.In order to verify the validity of the systems, and the justifiability of the fault-tolerant model, we established the satellite-based parallel system platform for fault-tolerant communication systems, and realized the 1553B bus fault-tolerant communication system client software. By using the application layer software and through the command line tool calling FTCP interface function, we verified the design of 1553B and PCIExpress bus fault-tolerant communication systems respectively. At last, the reliability of different strategies, protocol cost, the time of fault recovery are analyzed and compared.