Study On Multi-Tenant Data Migration Based On Metadata-Nodes

As cloud computing develops well, software as a service (Software as a Service) has attracted increasing attention which becomes a new state of software service. In order to let independent software vendors (ISV) to concentrate on developing the core business application, more convenient, flexible, fast to build a SaaS system, thus platform as a service (PaaS) concept develops. The PaaS platform to develop SaaS applications can hide many of the bottom implementation technology, has become the cloud computing market’main stream.The application of SaaS can provide services to multiple tenants at the same time, and every tenant can have one user or more than one, the infrastructure platform provides the service, but scalability brings a huge challenge. In this paper, we use the features of SaaS application, multi tenant to metadata driven shared data storage architecture as the underlying support, supplying the SaaS model mapping mechanism, smooth mapping application for metadata corresponding level, ensure that the SaaS application development fast and transparent, the prototype system of sharing database sharing data storage mode of data mode to construction of multi node data. Using this schema to achieve the greatest degree of resource sharing of multi tenant and improve resource usage rate, and it can make most profit, cutting the cost of operation effectively.Currently, the rapid development of cloud computing, the SaaS application delivery platform data is changing from single node to the multi node, and the number of data platform tenants quantity grows rapidly, the tenant data is stored on multiple nodes, in the face of the vast data and metadata, visit to data concurrent frequently, a node has been impossible to satisfy the performance requirements. And because the tenant requires resources space is changing dynamically over time. In this case, the access to hot data on some data nodes may lead to overload, the tenant of SLA (Service-Level-Agreement) is not fullfilled. Effective data migration can greatly weaken this situation. Therefore, this paper proposes multi tenant data migration strategy for dynamic cloud multi node SaaS application; this method can identify characteristics of tenants. In order to guarantee the continuity of the tenants of the access to source data node and the target data node, put forward the "single double reading" mode, the tenant SLA satisfied. In addition, the expansion of the two stage of traditional submitted strategies to ensure consistency of source and target data node.Finally, evaluate the migration strategy of multi tenant cloud data dynamic verification of multi node SaaS application with experiment, comparing the two migration method, namely multi tenant data migration algorithm, stop replication algorithm. Mainly through the transfer time, transfer speed, tenants in the migration process access speed and three performance indexes, the experimental results demonstrate that the migration strategy put forwarded in this paper can greatly cut down the downtime of migration process, and does not bother with the migration process access speed.The main jobs of this thesis are showed in the following:(1) Put forward the SaaS application sharing database sharing multiple data node model table model. Based on the base table and an extended SaaS multi tenant data storage mode and the multiple data nodes on a single SaaS platform, according to the metadata node problem, we put forward multi data nodes model. The model selects metadata nodes as metadata management node. The framework of metadata, the SaaS platform on the multi data nodes model, metadata distribution strategies, including metadata partitioning strategy based on segmentation and placement strategy definition. And proposed the segmentation access count property adjustment algorithm based on dynamic load. The model can effectively solve the problem of a single metadata node failure, performance fault, and expansibility.(2) Migration strategy of multi tenant data in the cloud of multivariate data node is proposed. Design a change in migration framework detection work load. Propose a minimum total cost of tenant identification strategy, the strategy according to the order of priority in order to perform the migration of tenants, which can decrease the quantity of data a migration, migration operation will avoid hard disk resources exhausting, or finish up the network bandwidth, to ensure that the overall performance of SaaS platform, and finally puts forward "double read write mode," in order to ensure the continuity of the multi tenant data in the migration process access.