Research On Low Power Scheduling Algorithm For NC System

The NC system which is the core technology of NC machine tools is a typical real-time system. In recent years, the development direction of the NC system is high-speed and high-precision. Therefore, there are high requirements for the NC system. Not only the task must complete within its deadlines, but also the task must complete successfully. More importantly, to ensure the accuracy of the system, when the task begins to execute and it occurs error before its completion, the system must deal with this error rapidly to ensure the failed task which completes its execution successfully.With the improvement of performance and increasing of the functionality in a NC system, the energy consumption of the NC system increases. In addition, with the rapid development of the VLSI technology and the size of CMOS circuits become smaller, the energy consumption of the NC system becomes higher. The higher energy consumption results in the high calories and the high temperature, which affects the life of the processor. The high temperature leads to increasing the probability of the system failure.Therefore, it will affect the reliability of the system. In addition, the higher energy consumption leads to increasing the cooling costs of the system which increases the production costs and decreases the competitiveness of the products. Moreover, the higher energy consumption will result in a waste of resources and the environment will be affected. Therefore, the energy consumption has become the bottleneck of the NC system design.This thesis takes the NC system as the background and analyzes the characteristics of the NC tasks. It combines the classic real-time scheduling theory and low-power technology. To solve the increasingly serious problem of the energy consumption, the more targeted, lower power consumption and higher reliability of low-power real-time scheduling algorithms which are based on the existing low-power real-time scheduling algorithms have been proposed. The specific research contents are as followings:Firstly, we present a periodic task low-power scheduling(DSTRA) algorithm and a discrete frequency periodic task energy saving(DSTRA_AD) algorithm. The DSTRA algorithm focuses on the independent periodic task model and it is based on early deadline first scheduling policy with a time complexity of nO)(. It not only uses the slack time from the higher priority task which completes early, but also uses the slacktime from the lower priority task. In addition, it considers the general power model. The DSTRA_AD algorithm which is based on the DSTRA algorithm uses the adjacent speed policy to execute the tasks for the processor providing the discrete speeds condition.The pseudo-code of the algorithm is given and next the feasibility of this algorithm is analyzed. The performance of the proposed algorithm is tested by simulation experiments.Secondly, we present a dynamic low-power sporadic task scheduling(DSTLPSA)algorithm. It focuses on the independent sporadic task model and it can reclaim the dynamic slack time to adjust the processor speed and use the DPM technique to reduce the energy consumption. In addition, it assumes that tasks are independent. In fact, there exists resources sharing problem between tasks in a NC system. Furthermore, it assumes that the execution time of a task scales linearly with the speed of the processor.In fact, the execution time of a task doesn’t scale linearly with the speed of the processor. To solve these problems, we present a dynamic low-power task scheduling algorithm with resource constraints(DSTSASR) algorithm. It is based on EDF/DDM scheduling policy and takes the resources sharing problem into consideration. Moreover,it can reclaim the dynamic slack time by building the slack time queue and it takes the overhead of the processor speed transition into consideration.Thirdly, to meet the requirement of the mixed task scheduling in a NC system, a constant bandwidth server mixed task low-power scheduling(CBSMTLPSA) algorithm has been proposed. It is based on the constant bandwidth server and the EDF scheduling policy. In addition, it combines the DPM technique and the DVS technique to reduce the energy consumption of the system. The pseudo-code of the CBSMTLPSA algorithm is given and next the feasibility of this algorithm is analyzed. The simulation experiments is carried out to validate the performance of the proposed algorithm.Fourthly, we present a dynamic low power sharing resource(DLPSR) algorithm. It focuses on the periodic task model with resources constraints. It not only takes the speed which affects the system reliability into account, but also it can reclaim the dynamic slack time from the early completion task and use the unused time from the recovery task to reduce the energy consumption. Therefore, it can improve the system reliability and reduce the energy consumption of the system.