Design Optimization Of Performance Guaranteed Energy Aware Real-time Applications

With the continuous development of computer technology, embedded real-time systems have been widely used in all walks of life. It is accelerating the process of intelligent era and it plays a vital role in human society. Embedded devices are increasingly applied in industry and people’s daily lives. The features of battery driven and small size for embedded system make that energy is very scarce resources. Thereby, optimizing the energy consumption of embedded systems is a serious challenge. Embedded systems are usually applied to some key areas, such as aerospace, defense and military, so the reliability of embedded system is a very critical issue. Due to the rapid development of the Internet in recent years and the increasing applications which are combined embedded real-time systems and Internet, this makes the security issue in the embedded real-time systems is even more important.In embedded real-time system, it is very strict for the deadline of the task. Due to the length of time for tasks to perform often vary widely and it is uncertain. Therefore, to ensure real-time tasks, many scholars usually consider worst-case execution time as the actual execution time for stochastic tasks. However, the probability that stochastic tasks actually reach the worst-case execution time is small, so the consideration will cause a great waste of system resources. In this thesis, mainly for real-time tasks stochastic, optimizing task scheduling with taking probability ensure for deadline, energy consumption, reliability and safety into considerations, eventually ensuring the performance of embedded real-time applications.Firstly, we establish the model of stochastic task, instead of the traditional task model that it must guarantee the tasks finish before their deadlines, we only ensure the probability of the real-time tasks miss deadlines. Recursion and Monte Carlo simulation are proposed to solve the problem. We optimize system expected energy consumption without violation of the given probability of no-deadline violation.We optimize the expected energy consumption while taking into account the system expected reliability. Because the use of DVFS may damage the reliability of the system, this thesis further considers this factor and establishs stochastic task reliability model. Optimizing expected energy optimization system with reliability and deadline constraints. Due to the complexity of the problem, this thesis designs optimal energy algorithm based on dynamic programming and a fast binary search algorithm which is an approximation algorithm to solve it.Finally, based on probability of no-deadline violation guaranteed for stochastic tasks, we further explore the security issue for stochastic tasks. For this reason, we establish a safety-critical model for stochastic real-time tasks. We use encryption algorithm to encrypt critical data task and then establish a system vulnerability model. A security-driven random real-time task scheduling problem is formulated. In order to solve the problem model, using mathematical theory of clustering idea, we design an approximate approach based on clustering for solving the vulnerability optimization problem.