【作者】 李萌；

【导师】 刘锋；

【作者基本信息】 安徽大学 ， 计算机技术（专业学位）， 2021， 硕士

【摘要】 近年来,业务流程管理(BPM)在云计算和大数据环境下有着前所未有的发展,数据密集型应用程序中的大量中间数据集可以被存储在云端以供二次使用。在如互联网+广告投放等业务流程管理系统中,由于源数据规模较大以及业务执行过程中会产生海量的中间数据集,从而导致在云端数据集的计算和存储成本过高。为了更好地利用云计算环境下的存储资源,目前有许多关于BPM数据的存储策略被提出,但大多数存储方法主要侧重于降低存储的成本效益,而较少侧重于减少被删除数据生成过程所需的响应时间。在云端所有的资源使用都需要支付相应的费用,包括数据集的存储费用以及再生被删除数据集的计算费用。由于中间数据集的大小、生成时间以及使用频率都是不同的,而这些都是决定在云环境下对数据集是否进行存储的重要参考因素。如何判断这些中间数据集是否需要进行存储,进而减少云端中所需要的计算资源和存储资源,以及避免生成数据集的响应时间过长,是具有实际意义的问题。针对这些问题,本文以减少数据集的成本和响应时间为目标进行研究,展开的具体工作包括以下三方面:(1)本文基于时间影响最大化方法构建时间影响模型,有效计算每个数据集的时间影响值。提出了SHTD数据集存储策略,实验根据时间影响模型计算出的时间影响值,通过SHTD策略在为给定预算下选择和存储数据集。SHTD是一种动态策略,首先存储具有最高时间影响值的数据集,然后更新其派生数据集的时间影响值,再依次选择数据集进行存储。基于此动态存储,SHTD策略比传统方法更有效降低时间和成本。(2)为了提高SHTD策略的性能,本文提出了TIMOM(time influence multi-objective optimization modol)数据存储模型。TIMOM模型的建立基于SHTD策略和多目标优化算法NSGA-Ⅱ,通过将总成本和响应时间作为两个目标,以时间影响值和数据集的大小为参数,对数据集进行了非支配排序,并对排序好的数据集计算拥挤距离,由此得到一组较适合存储在云中数据集,以供重复使用。实验结果表明,本文所提出的TIMOM模型可以有效地提高互联网+广告业务流程管理的性能。(3)本文提出MOIM(multi-objective optimization importance modol)数据存储模型。基于数据集的重要性思想,本文先构建重要性计算矩阵,再采用多目标优化的思想定义中间数据集的偏序关系,最后通过计算数据集的非支配排序及计算适应度函数构建MOIM模型,在一定条件下减少时间成本和存储成本。对比实验表明MOIM模型存储策略可以有效减少存储资源和计算资源。综上所述,论文面向互联网+广告业务流程管理中云数据集存储领域具体实际应用,从存储部分中间数据集的角度出发,以减少数据集的计算成本和生成响应时间为目的,基于多目标优化算法研究并提出了TIMOM模型以及MOIM模型,从而提高业务流程管理效率,这些工作对云数据存储实际应用领域具有实际参考意义。更多还原

【Abstract】 In recent years,business process management(BPM)has made unprecedented development in cloud computing and big data environment.And a large number of intermediate data setting in data-intensive applications can be stored in the cloud for secondary use.In business process management systems such as advertising in internet plus,the calculation and storage cost of data setting in the cloud is too high due to the large scale of source data and the huge amount of intermediate data setting generated during business execution.In order to make better use of the storage resources in the cloud computing environment,many storage strategies about BPM data have been proposed,but most of the storage methods mainly focus on reducing the cost-effectiveness of storage rather than,reducing the response time required for the generation of deleted data.All resources in the cloud cost corresponding fees,including the storage cost of data sets and the calculation cost of regenerating deleted data sets.Because the size,generation time and use frequency of intermediate data sets are different,these are important reference factors to decide whether to store data sets in cloud environment.How to make the decision whether these intermediate data sets need to be stored,so as to reduce the computing resources and storage resources needed in the cloud and avoid the long response time of generating data sets,is a practical problem.To solve these problems,this thesis aims to reduce the cost and response time of data sets,and the specific work includes the following three aspects:(1)This thesis builds a time impact model based on the time impact maximization method,and effectively calculates the time impact value of each data set.SHTD data set storage strategy is proposed.According to the time impact value calculated by the time impact model,the data set is selected and stored under a given budget by SHTD strategy.SHTD is a dynamic strategy,which first stores the data set with the highest time impact value,then updates the time impact value of its derived data set,and then selects the data set for storage in turn.Based on this dynamic storage,SHTD strategy can reduce time and cost more effectively than traditional methods.(2)In order to improve the performance of SHTD strategy,this thesis proposes TIMOM(time influence multi-objective optimization modol)data storage model.TIMOM model is based on SHTD strategy construction and multi-objective optimization algorithm NSGA-Ⅱ.by taking the total cost and response time as two objectives,taking the time influence value and the size of data sets as parameters,the data sets are sorted in a non-dominated way,and the crowded distance is calculated for the sorted data sets,thus a set of data sets is obtained,which is more suitable to be stored in the cloud for reuse.Experimental results show that the TIMOM model proposed in this thesis can effectively improve the performance of advertising business process management in internet plus.(3)In this thesis,MOIM(multi-objective optimization import modol)data storage model is proposed.Based on the importance of data sets,this thesis first constructs the importance calculation matrix,then defines the partial ordering relationship of intermediate data sets by adopting the idea of multi-objective optimization,and finally constructs the MOIM model by calculating the non-dominated ordering of data sets and calculating the fitness function,thus reducing the time cost and storage cost under certain conditions.Comparative experiments show that the storage strategy of MOIM model can effectively reduce storage resources and computing resources.In summary,this thesis is oriented to the practical application of cloud data set storage in internet plus advertising business process management.From the point of view of storing some intermediate data sets,aiming at reducing the calculation cost and generating response time of data sets,based on multi-objective optimization algorithm,TIMOM model and MOIM model are studied and put forward,so as to improve the efficiency of business process management,which has practical reference significance for the practical application field of cloud data storage.更多还原