Analysis of Checkpointing for Real-Time Systems

Predictable performance in the event of failuresis of paramount importance in most safety critical real-timesystems. Various hardware as well as software fault-toleranttechniques are employed towards this goal among which checkpointingis a relatively cost-effective scheme. Since checkpointing schemesdepend on time redundancy, they could affect the correctnessof the system by causing deadlines to be missed. This paper providesexact schedulability tests for fault tolerant task sets undera specified failure hypothesis and employing checkpointing toassist in fault recovery. The effects of checkpointing strategieson task response time are analysed and some insights for optimalcheckpointing are provided. The emphasis here is on utilizingthis analysis as an off-line design support tool.