A Real-time Operating System Can Survive And Achieve

Credibility is defined as the trustworthiness of systems which provides certain services and it has become an important research topic. Operating system is an important component of a computer system and a main target of study on credibility as well. At present, credibility of a computer system is mainly enhanced via improving reliability and fault-tolerance.Besides, there is another method to enhance the credibility of a system. That is to use the theory of survivability. It is stressed that critical services and normal services are treated in different ways in a survivable system. The real-time operating system, which as an important component of a real-time system, has a significant and overall impact on the whole system's behavior. In order to fundamentally improve the survivability of the operating system, we need to design certain mechanisms to support it. Here what we are going to talk about is to design a survivable real-time operating system for single-processor real-time systems.By studying the theory of survivability that from the field of information network systems, we figure out the definition of survivability and how to measure it for real-time system. After that, we analyze, design and implement a task management module of a real-time operating system, in which we also use the method of redundancy and imprecise computing model. In this management module, tasks are divided into two different parts, which are critical and normal ones. We try to make sure the implementation of critical tasks while to maximize the throughput of the whole system. Also we will design a mechanism upon checkpoints technology for critical tasks which could be resorted to for critical tasks to use the resource of normal tasks when faults occurred during implement. This task management module is consisted of two main parts: the scheduling part and task backup-recovery support part.Task scheduling part mainly talks about how to decide the part of tasks that will be implemented in the system and what order that these tasks will be scheduled. Two parameters are used here: criticality and priority. The criticality is for the purpose of distinguishing between critical tasks and normal tasks. And the priority helps to decide the order of tasks while scheduling. The implementation of critical tasks is always the most important thing to be guaranteed in our design. Only based on this premise, normal tasks can be considered to be applied. We talk about the problem of which part of tasks will be implemented in the system in two situations. Binary search method is used to solve this problem for the approximate optimal results.Task backup-recovery process can be used for critical tasks while faults occurred when they are running. As the result, the critical tasks can be roll back to the nearest checkpoint and try to fix the fault by implementing again rather than starting completely from the very beginning. It can save time and other resources which is very important to a time-mattered system. So supporting roll-back mechanism is very necessary for operating system.Next, we aim to introduce how to implement this design that talks previously on a real-time system which uses a Linux operating system after analyzed the kernel, and try to transplant.Finally, we introduce some experimental analysis that conclude the basic aim of our design was done.