支持SaaS应用多维异构性能需求的云资源放置方法

A Cloud Resources Placement Method Supporting SaaS Applications with Multi-Dimensional and Heterogeneous Requirements

SaaS(Software as a Service)应用是以云计算资源为基础,以按需定制及按需付费的服务模式向用户提供云计算软件服务的应用系统.云中的SaaS应用一般为多层多节点部署的大型软件应用系统,对于云计算SaaS服务提供商来说,往往需要在云数据中心中同时快速交付和部署多个不同的SaaS应用,需要满足不同租户对于不同的SaaS应用多样化性能、网络、存储和操作系统需求,即多维异构的性能环境需求.因此,如何快速选择合适的云资源来部署大规模SaaS应用系统,满足大规模不同租户的多维异构性能需求,同时节省云服务提供商的成本,是实现SaaS应用敏捷交付部署的关键.传统的按照等级和供需的云资源匹配方法已经很难满足云数据中心大规模SaaS应用敏捷化交付部署要求.为此,提出一种基于图匹配的SaaS应用云资源放置方法,将大规模SaaS应用的个性化云服务放置问题映射为云资源节点拓扑图的子图查询匹配问题,即SaaS应用的多节点多维性能需求和云资源节点拓扑均表示为带多维属性标签的异构图,基于偏序异构图查询匹配方法得到一组满足用户需求的云资源节点集合,用于放置SaaS应用及其数据,从而实现大规模SaaS应用的敏捷化交付部署.实验结果表明该方法能有效提高大规模复杂SaaS应用多维异构云资源放置的执行效率.

Software-as-a-Service（SaaS）is a new software delivery model that provides on-demand customization and payment for tenants based cloud platform.SaaS has drawn considerable research attention for its capability of transferring software goods for services in light of the Internet based platforms.Individual software vendors release resources and services to tenants by deploying the SaaS applications accordingly.SaaS applications in cloud are generally large scale multi-layer software application systems,which deployed in multiple nodes and are very important to tenants.On the other hand,SaaS service providers wish to reduce the total cost and gain more benefit.To this end,SaaS service providers require to quickly deliver and place multiple diversified SaaS applications in the cloud data center to meet the multi-dimensional and heterogeneous requirements,such as performance,network utilization,storage capability and operating system requirements.As a result,it is essential to achieve the SaaS application agile placement to quickly select appropriate cloud resources to place large scale SaaS application system to meet the needs of large scale heterogeneous performance requirements of different tenants and hence reduce the cost of cloud service providers at the same time.Therefore,the key point of placing SaaS applications in an efficient way lies on choosing the appropriate cloud resources to deploy the SaaS applications.Traditional cloud resources matching methods based on Service Level Agreement SLA and supply demand have been difficult to adapt to the requirements of agile placement of large scale SaaS applications in cloud for the complicated requirements of large-scale heterogeneous performance.In this paper,the strategy of placing SaaS applications based on graph matching theory is proposed which models the SaaS applications placement problem as the subgraph matching problem of the cloud resource graph.The computing resources,data resources and their inherent connections in the cloud are mapped into an isomerism graph with multi-dimensional attribute labels.In this graph,vertices represent the cloud resources required by SaaS applications,while edges represent topological relationships between cloud resources.The attribute values on vertices represent the performance requirements of SaaS applications for cloud resources.After constructing the above graph,an algorithm which contains four steps is presented.The first step is to mine the frequent subgraphs which satisfy a certain threshold in the cloud services resource graph.In the next step,the request graph is cut using frequent subgraphs.Then the subgraph set of the request graph is obtained.The third step is to filter the subgraph set of the request graph.In this process,the algorithm dexterously uses the coordinate division method to improve the efficiency of the filtering operation.In the final step,the candidate set is merged to reduce the complexity of the problem and solve the SaaS applications placement problem quickly and accurately.A set of cloud resources that meet the provider＇s requirements can be eventually obtained by using the partial order relation isomerism matching method,which is employed to agilely place SaaS applications and data.The extensive experimental results suggest that the proposed method illustrates significant improvement in the efficiency of multi-dimensional heterogeneous cloud resource placement strategy in complex SaaS applications.