多租户数据放置与访问研究

发布时间：2018-05-05 04:37

本文选题：SaaS + 放置与访问　；参考：《山东大学》2014年硕士论文

【摘要】：伴随多租赁理念在业界的推广应用,基于SaaS模式的应用已经成为一种高效、先进的业务应用解决方案。单实例多租赁(Single Instance Multi-tenancy)的应用模式体现出其低费用,低维护,高效应用的特点；SaaS应用的软件提供商,为了能够提供更好的软件服务,对部署的应用服务SLA有了更高的要求,而云计算的出现,恰恰能够提供高可用的系统应用和高扩展的应用部署和数据服务。面对共享架构共享存储的云数据SaaS应用,多租户数据被放置到云中多个数据节点上,而随机或者不合理选取云数据节点进行放置,容易引发应用数据节点的负载不均衡,同时也可能会增加租户数据访问及副本一致性更新的代价；针对云中多租户数据访问,结合多租户和云存储特点,如何实现多租户数据应用请求访问与合理调度策略,从而来降低节点负载并保持节点间负载均衡,实现系统的高效稳定运行。面对这些问题,本文设计了多租户云平台系统存储架构模型,在此存储模型的基础之上,提出了多租户数据放置与访问的目标与解决方案,主要贡献如下： 1、针对随机或不合理的云数据放置,导致数据节点负载不均衡问题,设计多租户数据放置与访问算法,获得最优的多租户数据放置与访问策略。通过引入多租户数据放置权重函数,构建多租户数据放置与访问模型,利用图算法理论,提出了基于节点网络距离和数据负载的数据放置与访问算法,实现了基于图的最优完全图的放置与访问算法,并对算法进行正确性和复杂性的分析,保证了数据的合理放置和节点的负载均衡,减小租户数据访问及副本一致性更新的代价。 2、针对云中多租户数据访问,设计多租户数据访问请求处理功能与流程,实现了面向云中多租户的数据虚拟化访问。通过提出面向云中多租户的数据请求访问模型,设计了云中多租户数据的访问请求目标,针对不同租户类型和访问数据类型的,详细描述云中多租户数据访问处理流程,实现根据多租户元数据、业务数据等不同数据类型的多租户数据访问请求处理。 3、面对云中多数据节点请求访问与数据不一致性问题,利用多租户数据放置与访问模型和算法,提出根据节点负载状态的动态调度访问模型,实现多租户数据的高效访问；借鉴Paxos算法,结合多租户元数据和多租户特点进行改进,引入元数据描述队列的数据结构,解决云中多租户数据可能出现的数据不一致的问题,保证SaaS应用系统的可用性。结合文中提出的多租户数据放置与访问模型与算法,本文通过实验评估证明了多租户数据放置的正确性和动态调度访问的有效性；文中提出的模型与策略在一定程度上不仅可以提高租户之间的共享程度、改进数据放置的负载均衡,而且可以减轻数据副本更新的时间,减少租户访问数据的代价,为云中SaaS数据管理与应用提供了参考与帮助。
[Abstract]:With the application of multi - tenancy concept in the industry , the application of the SaaS model has become an efficient and advanced business application solution . The application mode of Single Instance Multi - tenancy ( Single Instance Multi - tenancy ) shows its low cost , low maintenance and high efficiency application .
SaaS - based software providers , in order to provide better software services , have higher requirements for deployed application service SLAs , while cloud computing is emerging to provide highly available system applications and highly scalable application deployment and data services .

Facing the cloud data SaaS application shared by the shared architecture , the multi - tenant data is placed on a plurality of data nodes in the cloud , and the cloud data nodes are randomly or unreasonably selected to be placed , the load of the application data node is easily caused to be unbalanced , and meanwhile , the cost of the tenant data access and the copy consistency update may be increased ;
Aiming at multi - tenant data access in the cloud , combining the characteristics of multi - tenancy and cloud storage , how to realize multi - tenant data application request access and reasonable scheduling strategy can be realized , so that the node load can be reduced and the load balance among nodes is maintained , and the efficient and stable operation of the system is realized .

In the face of these problems , this paper designs a multi - tenant cloud platform system storage architecture model , based on this storage model , puts forward the target and solution of multi - tenant data placement and access , and the main contribution is as follows :

1 . Aiming at random or unreasonable cloud data placement , the data node load imbalance problem is caused , and the multi - tenant data placement and access algorithm is designed to obtain the optimal multi - tenant data placement and access strategy .

By introducing a multi - tenant data placement weight function , a multi - tenant data placement and access model is constructed . Based on the theory of graph algorithm , a data placement and access algorithm based on node network distance and data load is proposed , the placement and access algorithm of the optimal complete graph based on the graph is realized , the correctness and complexity of the algorithm are analyzed , the reasonable placement of the data and the load balance of the nodes are guaranteed , and the cost of the tenant data access and the copy consistency update is reduced .

2 . For multi - tenant data access in the cloud , the multi - tenant data access request processing function and the process are designed , and the data virtualization access for multi - tenant in the cloud is realized .

According to the data request access model for multi - tenant in the cloud , the access request target of multi - tenant data in the cloud is designed , and the multi - tenant data access processing flow of multi - tenant data in the cloud is described in detail aiming at different tenant types and access data types , and multi - tenant data access request processing according to different data types such as multi - tenant metadata and service data is realized .

3 , facing the problem of request access and data inconsistency of the multi - tenant data node in the cloud , using the multi - tenant data placement and access model and the algorithm , and putting forward the dynamic scheduling access model according to the node load state to realize the efficient access of the multi - tenant data ;
This paper introduces the data structure of multi - tenant metadata and multi - tenancy , introduces the data structure of metadata description queue , and solves the problem of inconsistent data of multi - tenant data in the cloud , and ensures the availability of SaaS application system .

Based on the multi - tenant data placement and access model and algorithm proposed in the paper , the correctness of multi - tenant data placement and the validity of dynamic scheduling access are proved through experimental evaluation .
The proposed model and strategy can not only improve the sharing degree among tenants , improve the load balance of data placement , but also reduce the time of data copy update , reduce the cost of tenant access data , and provide reference and help for SaaS data management and application in the cloud .

【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】