【作者】 张涛；

【导师】 张立臣；

【作者基本信息】 广东工业大学 ， 软件工程（专业学位）， 2020， 硕士

【摘要】 随着分布式、大数据、云计算等信息技术以及传感器和嵌入式等硬件技术的快速发展,使得CPS技术在解决工业中实际应用场景的问题有了多种多样的解决方案。复杂工业中大数据下的CPS系统任务调度过程中容易因为数据高并发、资源异构、任务性质复杂造成系统负载不均衡,导致在任务调度过程中因任务量积累的负载增加无法进行均衡,直至系统宕机。即使目前有许多云计算调度算法能够很好的解决分布式集群系统中的负载均衡问题,但在CPS系统具体情形下的复杂因素,只考虑系统负载或者只考虑任务的属性的调度算法仍容易出现系统不稳定的状态,造成调度整体效率下降的问题。1.针对上述问题,本文从系统的整体角度进行思考和研究,同时考虑系统负载和任务属性对任务调度的影响,提出一种以性能指标调控系统负载和任务执行状态,确保系统以最佳状态执行任务的前提下,及时降低系统处于不稳定状态的风险的任务调度方案。其包括如下几个方面:2.深入了解CPS系统任务调度算法的背景,如云计算、大数据,分析其特性。研究当前任务调度的发展状况,分析各种算法的优缺点。3.研究排队论的原理,分析不同排队模型的组成形式,选择符合复杂工业下的CPS系统结构的排队模型作为本文提出的调度算法的运行场景。研究分析该模型下各项指标参数的计算方式和相关性,并对排队模型进行系统建模,结合排队论提出基于排队论动态反馈CPS任务调度策略模型(QTTS算法)。经过定量分析后,为了优化算法在某些方面的不足以及提高CPS系统对不同价值的任务调度效率,提出一种基于任务服务价值多级优先级排序算法与任务反馈调度算法融合,进一步提高调度算法的性能和适应度。本文利用Cloud Sim Toolkit 3.0云计算仿真平台对本文的调度算法进行仿真实验,其实验结果证明本文的提出的调度算法论与传统算法相比,任务的响应时间快,负载平衡能力强以及对任务价值调度效率高,论证了本文提出的调度策略的可行性和运作性能,最后总结算法的优缺点,以及下一步的研究方向。更多还原

【Abstract】 With the rapid development of distributed,big data,cloud computing and other information technologies,as well as sensors and embedded hardware technologies,CPS technology has a variety of solutions to solve the problems of practical application scenarios in industry.In the complex industry,the task scheduling process of the CPS system under big data is prone to unbalanced system load due to high data concurrency,heterogeneous resources,and complex task nature.As a result,the load accumulated during the task scheduling process cannot be balanced due to the increase in load accumulated by the task amount The system is down.Even though there are many cloud computing scheduling algorithms that can solve the load balancing problem in distributed cluster systems well,the complex factors in the specific situation of the CPS system,scheduling algorithms that only consider the system load or only consider the attributes of tasks are still prone to appear The unstable state of the system causes the overall efficiency of scheduling to drop.In view of the above problems,this paper considers and studies from the overall perspective of the system,and at the same time considers the impact of system load and task attributes on task scheduling,proposes a performance index to regulate the system load and task execution status to ensure that the system performs tasks in the best state Under the premise of,the task scheduling scheme to reduce the risk of the system in an unstable state in a timely manner.It includes the following aspects:1.In-depth understanding of the background of the CPS system task scheduling algorithm,such as cloud computing,big data,and analyze its characteristics.The current development status of task scheduling,analyze the advantages and disadvantages of various algorithms.2.Study the principle of queuing theory,analyze the composition of different queuing models,and select the queuing model that conforms to the CPS system structure in complex industry as the running scenario of the scheduling algorithm proposed in this paper.Study and analyze the calculation method and correlation of each index parameter under the model,and systematically model the queuing model.Combined with the queuing theory,a dynamic feedback CPS task scheduling strategy model based on queuing theory is proposed(Queuing Theory Task Scheduling).After quantitative analysis,in order to optimize the deficiencies of the algorithm in certain aspects and improve the efficiency of the CPS system for scheduling tasks of different values,a fusion of multi-level priority ranking algorithm based on task service value and task feedback scheduling algorithm is proposed to further improve the scheduling algorithm Performance and fitness.3.This paper uses the Cloud Sim Toolkit 3.0 cloud computing simulation platform to perform simulation experiments on the scheduling algorithm in this paper.The experimental results prove that the proposed scheduling algorithm theory in this paper is faster in task response time,stronger load balancing ability,and more effective The high efficiency of task value scheduling demonstrates the feasibility and operational performance of the scheduling strategy proposed in this paper,and finally summarizes the advantages and disadvantages of the algorithm,as well as the next research direction.更多还原