Research On The Optimization Of Energy Consumption For Real-time System Based On DRT

With the rapid development of information technology,the real-time system has been widely used in various fields.However,the continuous improvement of the system performance also brings the energy consumption problem which could not be ignored.Low energy consumption is an important indicator to measure how well the real-time system is.Due to the increasingly prominent contradictions between the slow development of battery technology and the growing demand of the performance,the increasing cost of encapsulation and cooling,the adverse effects caused by high temperature on the stability and reliability of real-time system and the theme of the times that advocates the concept of low-carbon and environmental protection,the optimization of energy consumption for real time system has become an urgent problem to solve.On the other hand,in real-time scheduling,with the purpose of resolving the contradiction between the deadline of all tasks and efficient use of system resources,we usually establish task model and adjust the parameters of the execution time to solve this issue.Compared with the other task models and the task model newly proposed-the Digraph Real-Time Task Model(DRT)has the advantages of both high expressiveness which means describing the diversity of tasks and high efficiency in analysing tasks,and it has played an important role in the research field of the real-time system.According to the energy consumption problem of the real-time system and the advantages of DRT model stated above,a new algorithm for the optimization of energy consumption for real-time system based on DRT is proposed in this thsis.The core idea of the algorithm is to calculate the maximum execution time of each job of a real-time task which can be improved to make its corresponding demand bound function as close as possible to the given target demand bound function and then guarantee the schedulability of task and reduce the energy consumption of the system when improving execution time through reducing the frequency.This thesis assumes that each real-time task has a priority,and the priority reduces from high to low according to the energy consumption decreased of the increase of unit execution time.In this thesis,we use the dynamic voltage scaling technique and establish a processor model of discrete adjustable frequency which is more close to the actual system and the corresponding discrete energy consumption model.At last,through a large number of comparative experiments,the thesis analyzes the results of the experiments and proves the feasibility and efficiency of the algorithm.