This paper addresses the fault-tolerant performance of real-time dual-computer embedded systems. The article first emphasizes the importance of real-time and reliability in various fields and points out that improving fault-tolerant performance is a crucial topic. Based on the Markov chain algorithm, the study optimizes the fault-tolerant integration method for real-time dual-computer embedded systems. By constructing a model of Markov algorithm and using the deadline of the task as a benchmark, the passage and transfer probabilities of faults are calculated. The article also provides algorithmic proofs of fault-tolerant control of Markovian jump systems and calculates their stability levels. The results show that the fault passage rate of the system increases as the number of complex tasks increases, e.g., when the number of complex tasks is 4, the passage rate can reach 90%. In addition, in the scheduling test, it was found that the schedulability of the system increases with the increase in the number of processors. When the number of processors reaches 5, the system's schedulability is 43%. In conclusion, the system fault tolerance optimization method based on Markov algorithm proposed in the article can effectively improve the reliability and fault tolerance of the system.