基于云平台的业务流程引擎任务调度算法研究
发布时间:2019-03-28 18:18
【摘要】:近年来,随着信息技术的快速发展,公司企业越来越重视部门、组织间的沟通效率,业务流程管理(Business Process Management,BPM)技术可以为跨组织、跨部门业务集成等方面提供灵活有效的管理方式,因此,业务流程管理系统在企业中应用越来越广泛。但是随着企业的发展,业务的不断增多,许多企业在实施BPM技术的过程中都面临耗费大、缺乏扩展性以及响应效率低的问题。云计算的特点恰好为企业在实施业务流程管理过程中遇到的问题提供解决方案。本文将结合BPM技术和云计算技术来解决传统BPM面临的诸多问题。通过对云工作流任务调度、云服务相关的QoS、负载反馈相关技术和理论的研究后,本文主要从以下几点进行研究:首先提出基于云平台的分布式业务流程引擎模型,该模型采用主从架构,由任务监控节点和任务服务节点组成,运用Hadoop平台中的分布式文件系统来存储流程定义文件;然后在该模型的基础上进行任务调度算法的研究,提出基于QoS的任务分配算法和基于负载反馈的延迟调度算法。基于QoS的任务预调度算法是指通过用户对流程服务的QoS需求,把QoS需求考虑在任务分配过程中,运用遗传算法计算出任务的分配策略,提高整个系统的任务吞吐量;同时为了尽可能保证每个业务流程任务在服务节点执行时可以获得充足的资源,又提出基于负载反馈的延迟调度算法,重点研究调度时机的选择,优化任务在服务节点的执行次序,将最迫切的请求优先调度。最后,通过实验对比运用遗传算法和顺序分配算法在相同流程定义的情况下全部任务的完成时间,并对结果进行分析总结,实验结果表明,运用遗传算法生成的分配策略要比顺序分配执行的时间更短。同样,对运用负载反馈的延迟调度算法进行实验数据分析,实验表明该算法能够有效的平衡资源,提高资源利用率。
[Abstract]:In recent years, with the rapid development of information technology, companies increasingly attach importance to departments, inter-organizational communication efficiency, business process management (Business Process Management,BPM (business process management) technology for cross-organization. Cross-departmental business integration provides flexible and effective management methods, so business process management systems are more and more widely used in enterprises. However, with the development of enterprises and the increasing of business, many enterprises are faced with the problems of high cost, lack of scalability and low response efficiency in the process of implementing BPM technology. The characteristics of cloud computing provide solutions to the problems that enterprises encounter in the process of implementing business process management. This article will combine BPM technology and cloud computing technology to solve many problems that traditional BPM faces. After studying the technology and theory of cloud workflow task scheduling and cloud service-related QoS, load feedback, this paper mainly focuses on the following aspects: firstly, a distributed business process engine model based on cloud platform is proposed, which is based on cloud platform. The model adopts master-slave architecture, which consists of task monitoring node and task service node. The distributed file system in Hadoop platform is used to store process definition files. Then the task scheduling algorithm based on this model is studied and the task assignment algorithm based on QoS and the delay scheduling algorithm based on load feedback are proposed. The task pre-scheduling algorithm based on QoS is to calculate the task allocation strategy by using genetic algorithm to improve the task throughput of the whole system by taking the QoS requirement into account in the process of task assignment through the QoS requirement of the user to the process service. At the same time, in order to ensure that every business process task can obtain sufficient resources when the service node is executed, a delay scheduling algorithm based on load feedback is proposed, and the selection of scheduling timing is emphasized. Optimizes the execution order of tasks at the service node and prioritizes the most urgent requests. Finally, the completion time of all tasks is compared by genetic algorithm and sequential assignment algorithm under the same process definition, and the results are analyzed and summarized. The experimental results show that: The allocation strategy generated by genetic algorithm takes less time than sequential allocation. In the same way, the experimental data of the delay scheduling algorithm based on load feedback is analyzed. The experimental results show that the algorithm can effectively balance the resources and improve the utilization rate of the resources.
【学位授予单位】:哈尔滨工程大学
【学位级别】:硕士
【学位授予年份】:2016
【分类号】:TP301.6;TP393.09
[Abstract]:In recent years, with the rapid development of information technology, companies increasingly attach importance to departments, inter-organizational communication efficiency, business process management (Business Process Management,BPM (business process management) technology for cross-organization. Cross-departmental business integration provides flexible and effective management methods, so business process management systems are more and more widely used in enterprises. However, with the development of enterprises and the increasing of business, many enterprises are faced with the problems of high cost, lack of scalability and low response efficiency in the process of implementing BPM technology. The characteristics of cloud computing provide solutions to the problems that enterprises encounter in the process of implementing business process management. This article will combine BPM technology and cloud computing technology to solve many problems that traditional BPM faces. After studying the technology and theory of cloud workflow task scheduling and cloud service-related QoS, load feedback, this paper mainly focuses on the following aspects: firstly, a distributed business process engine model based on cloud platform is proposed, which is based on cloud platform. The model adopts master-slave architecture, which consists of task monitoring node and task service node. The distributed file system in Hadoop platform is used to store process definition files. Then the task scheduling algorithm based on this model is studied and the task assignment algorithm based on QoS and the delay scheduling algorithm based on load feedback are proposed. The task pre-scheduling algorithm based on QoS is to calculate the task allocation strategy by using genetic algorithm to improve the task throughput of the whole system by taking the QoS requirement into account in the process of task assignment through the QoS requirement of the user to the process service. At the same time, in order to ensure that every business process task can obtain sufficient resources when the service node is executed, a delay scheduling algorithm based on load feedback is proposed, and the selection of scheduling timing is emphasized. Optimizes the execution order of tasks at the service node and prioritizes the most urgent requests. Finally, the completion time of all tasks is compared by genetic algorithm and sequential assignment algorithm under the same process definition, and the results are analyzed and summarized. The experimental results show that: The allocation strategy generated by genetic algorithm takes less time than sequential allocation. In the same way, the experimental data of the delay scheduling algorithm based on load feedback is analyzed. The experimental results show that the algorithm can effectively balance the resources and improve the utilization rate of the resources.
【学位授予单位】:哈尔滨工程大学
【学位级别】:硕士
【学位授予年份】:2016
【分类号】:TP301.6;TP393.09
【相似文献】
相关期刊论文 前10条
1 鲁晓霞;周忠和;;云计算中主任务调度算法的优化研究[J];计算技术与自动化;2011年04期
2 禄乐滨,刘明东;一种基于函数的多任务调度算法[J];空军工程大学学报(自然科学版);2000年02期
3 阮幼林 ,刘干 ,朱光喜 ,卢小峰;一个基于复制的相关任务调度算法[J];小型微型计算机系统;2005年03期
4 杨斌;张建军;;一个新的基于通信竞争的任务调度算法[J];计算机工程与应用;2007年33期
5 胡同福;王文生;谢能付;;设备网格中的任务调度算法[J];计算机工程与设计;2008年12期
6 周艳慧;张凯;;新的分布式任务调度算法[J];计算机系统应用;2008年10期
7 薛继伟;姜波;刘庆强;王征;;基于能力感知的人机任务调度算法[J];计算机工程;2009年19期
8 曹晓磊;程东年;黄万伟;;基于离散时间距的在线可重构任务调度算法[J];小型微型计算机系统;2010年10期
9 韩晓亚;汪斌强;黄万伟;王保进;;采用配置完成优先策略的可重构任务调度算法[J];小型微型计算机系统;2012年03期
10 杨丽;武小年;商可e,
本文编号:2449113
本文链接:https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2449113.html