Design Of A Profile Based Checkpoint In Software Fault Tolerance

The on-board computers work in outer space full of space radiation. The high-energy cosmic rays is one of the most important effectors which will lead to the SEU(Single Event Upset) of memory device. SEU will trigger a hardware transient error, which makes software failure and error probability exponentially. To meet the demand for on-board computers of high reliability, hardware and software reinforcement reinforcement indispensable.Most embedded real-time operating sysytem will provide a set of built- in error detection recovery mechanisms to deal with SEU. Restart the task is one of main means, but it will cause serious waste of system resources. Therefore, develop a mechanism of software reinforcement possesses both the reliability and real-time is imperative.In this paper are as follows:Firstly, analysis the working environment of on-board computer and the reliability needs deeply. Then classify and compare the current fault-tolerant technology, especially for checkpoint software reinforcement mechanism. Combined with probability theory, analysis the feature of transient hardware errors caused by SEU, we put forward profile based model and PBC algorithms.Compared with conventional checkpoint, the mechanism has the following two advantages: First, in the same piece o f code and the same number of checkpoints, profile based checkpoint significantly shorter mean time to recovery; Second, through a user guide, dynamic optimization checkpoint insertion position can quickly adapt to changes in the space environment.Finally, using a BM3803 processor and VxWorks embedded operating system to build experimental platform, we compare the efficiency of program recovery of profile based checkpoint and conventional checkpoint mechanism. Experimental results show that the profile based checkpoint has obvious advantages over conventional checkpoint mechanism.