| In the context of the commercialization of 5G and the rapid development of Internet technology,emerging applications such as augmented reality/virtual reality and webcasts are gradually appearing in people’s lives,which put forward higher requirement for network latency and computing capacity and other aspects.In order to guarantee the quality of service,the mobile computing model has changed from traditional cloud computing to mobile edge computing(MEC).The main feature of MEC is to deploy network resources closer to the edge of the network to meet the needs of delay-sensitive and computation-intensive services.With the rapid growth of mobile data traffic,there will also be more intense competition among users for resources,and how to reasonably schedule tasks to meet the service requirements with limited resources is a major challenge.In this thesis,this problem is studied and analyzed from two scenarios of single server and multi-server.The main work is as follows:Firstly,aiming at the problem of queuing delay of tasks in a single server,in order to meet the delay requirements of different tasks,an improved DRR algorithm based on sub-period is designed.Compared with the classical DRR scheduling algorithm,the improved algorithm divides a complete period into multiple sub-periods,and the quantum of each queue will be consumed in multiple sub-periods instead of one time,which can reduce the waiting delay in the worst case.In addition,in order to provide better service for delay-sensitive tasks,the waiting queue is set as a high priority queue,and the other queues are set as normal priority queues.Every time a normal priority queue is scheduled,the high priority queue will also get a chance to be served,which can greatly reduce the waiting delay of such tasks.Then,the service curve under the improved algorithm is derived step by step based on network calculus,and the service curve expressions under different priority queues are given.In order to make a more comprehensive analysis of the delay under the algorithm,two different arrival curve models of single leaky bucket and dual leaky bucket are considered,and the delay bounds under different configurations for the two models are also calculated.The improved algorithm is compared with the DRR algorithm,and the effectiveness of the proposed algorithm is proved.Secondly,aiming at the task scheduling problem in multi-task and multi-server scenario,in order to meet the task requirements and reduce the cost as much as possible,an improved algorithm based on artificial fish swarm algorithm(AFSA)was designed.Most of the existing researches do not consider the queuing delay in this scenario.In this thesis,we first apply the improved DRR algorithm based on sub-cycle to each server in this scenario,and take the queuing delay into account when establishing the delay model.And the energy consumption and task characteristics are analyzed,and the corresponding model is established.Then,with the goal of minimizing the weighted sum of delay and energy consumption,the solution formula and related constraints were constructed.In the solution of the optimization problem,based on the AFSA,the parameter adaptation and elite retention strategy are used to improve it,which aims to ensure that the search range is large enough and the optimization accuracy is high,and the excellent characteristics can be retained in the process of iteration.Finally,the convergence of the algorithm is analyzed by simulation,and the influence of different parameters on the performance of the algorithm,such as the number of tasks,calculation density and bandwidth size and so on,is also analyzed,and compared with particle swarm optimization(PSO),genetic algorithm(GA)and other algorithms,the performance of the proposed algorithm is verified.In this study,the above-mentioned two improved algorithm are simulated and analyzed,and the performances are compared.It is verified that the proposed algorithms can effectively improve the performance. |
PDF Full Text Request