Meet The Survivability Of The Research And Implementation Of Collaborative Checkpoint

Many real-time systems are running as the core of critical safety system. Purpose of survivability research is to protect the important imformation and essential services when system can't offer complete standard services under attacks, faults or failures. To estimate system survivability, first, it needs to define the specifications of acceptable system services with multiple alternate forms, to offer proper level protection for critical and general services. Second, to analyze the services, quantify the service value and service probability.The high dependability and strict time condition of real-time system must be guaranteed. Fault tolerance is typically achieved in real-time systems through a combination of real-time schedule and checkpoint mechanism. Checkpoint can reduces the time of recovery process to improve system availability. The two main approaches are system level and user level checkpoint which both have advantages and disadvantages. But the failure distributiones assumed by most of researches are not fit for realistic environment.First, this dissertation presents the current state survey of survivability research, fault-tolerant real-time schedule algorithm and checkpoint implements. Then it analyzes adventages and disadventages of system and user level checkpoint, and introduces the cooperative checkpoint which combines the adventage of checkpoint both level. Second, it researches the real-time fault-tolerant schedule based on cooperative checkpoint. To meet survivability demands, it analyzes task set feasibility condition under checkpoint for choosing proper checkpoint strategy under different system performance.Third, this dissertation desingnes and implements the checkpoint system based on BLCR. The system is composed of kenerl module, user module, and watch prediction module. It can choose checkpoint setting based on system perfomance and failure imfomation. And the experimental simulated in linux environment presents that even deleasing the number of checkpoint, cooperative checkpoint can get similar performance compare to periodic checkpoint. So cooperative checkpoint reduces checkpoint cost and enhances system fault-tolerant capability.