Design And Implementation Of Task-level Recomputation In On-Board Parallel Computer System

Nowadays, the real time embedded system is widely used especially in some domain with high demanding of reliability, such as small satellite and flight control systems etc. The research of the embedded fault-tolerant in real-time system gradually become a hot spot.Usually, application programmers have a great knowledge of their application, so the efficiency of implement fault-tolerance in the application level will be relatively high.. However, there are two problems in fault tolerance in the application layer. On the one hand, fault tolerant consideration overburden application developers in that application developers consider not only how to achieve the functions of application software, but also how to make application fault-tolerant. This increases the workload and complexity of application development. More seriously, it raise the possibility of error happening. On the other hand, during application development, the same fault-tolerance mechanism always put into practice repeatedly, which is not consistent with the rule of improving software reusability, reducing costs of development and improving reliability in software development. Therefore, the design of a certain universal fault-tolerant software modules is meaningful.In this paper, we aim at improving the reliability of on-board parallel multi-processing computer. On the bases of in-depth analysis of single particle effects on software applications and drawing on the ideological N-Version programming fault-tolerant technology, we present a task level fault-tolerant technology. Through in-depth analysis of RTEMS kernel, we build a task-level complex computation support layer, giving programmers a easy and cost-effective way to develop application. The main contributions of this paper are shown below. First, solve the problem of how to distribute copy tasks in the multi-processing system. Second solve the problem of the preservation, restoration and renewal of task parameter. Third, present the technique of error detection and error handle. Task-level complex computation support layer blend the fault tolerant technique of task-level complex computation mechanism with the operating system, without making system inefficient or overburdening application programmers, it improves the reliability of application task.