Critical Techenique Research On Schema Evolution For SaaS

In recent years, software is becoming networking, the trend of making software delivery model, application model, product form and business model have had a great changes. Software has become a service (Software as a service, SaaS), that is, people with " Use instead of own "way to use the software, users no longer need to install any software local. On-demand, no maintenance, online updates, and free or pay-as-use software as a service (SaaS) rental business models are gaining increasing attention and recognition. SaaS applications with the "single instance multi-tenancy" feature, that same application instance on the same hardware and software platform for multiple tenants to provide personalized services. In order to fully share the software and hardware resources, it has centralized data management, application of the same data structure not only to be shared by multiple tenants, while tenants have to meet different data models of their customization sufficient requirements and does not affect other tenants. When the tenant scale to a certain extent, the application of the accumulated data the user requests the load capacity and the growth rate is amazing, this phenomenon of high-load applications, SaaS will upgrade a significant impact.Upgrade applications, database schema evolution is an essential step, schema evolution can be divided into two steps:pattern matching and data conversion. Pattern matching refers to the source model and target model are identified under the premise of the source model to target model from the optimal mapping, which gives the source model and target model of each element of the corresponding mapping of elements, and data conversion steps calculation provides a foreshadowing. Relational database schema in the data conversion is actually a set of SQL operations, the source model in the implementation of these actions can evolve to the target model. Evolution of the static model is to apply the services stop, once the source mode to perform all of the data conversion operation, and then evolved to the target model. SaaS application upgrade, legacy application co-exist with the new, different versions of applications designed for optimal efficiency of the database access to different modes, as well as multiple versions of the database application maintenance independent of each other will lead to data consistency problems, so should only maintain a database multi-version application for access. If the way the evolution of static mode direct evolution of the source model to the target mode will take a long time, resulting in long-term services of free time and can not guarantee that all users access to the overall efficiency.In this paper, the method of gradual evolution model, SaaS applications in determining the most efficient access to new and old versions of the model (source model and target model), after the software upgrade process to maintain only an intermediate model, and applied according to the load on different time the amount of data changes in the distribution of its gradual evolution, it evolved gradually from the source model to the target model to ensure the minimum consumption of resources during the upgrade. This mode of evolution strategy of gradual evolution in static mode on the basis of evolved, it first pattern matching and data conversion using the relationship between model and target only the source model as input, the use of SF pattern matching algorithms and static data conversion strategy to get the static mode evolution pattern matching and data conversion when the results; after the help of genetic algorithm, described the global optimum progressive mode of evolutionary algorithms. The main work:An application in the SaaS version upgrade in the context of proposed applications of a more realistic model of gradual evolution framework. Data that is generated automatically based on the conversion steps, the steps assigned to each point on the evolution of the time evolution of each point to a time when the appropriate action, in order to achieve optimal resource consumption.2 This paper pattern matching, data conversion process into account, the source model and target model are identified under the premise is given a complete set of data conversion process automatically calculated steps. SF is the first pattern matching algorithm using the original model to calculate the mapping to the target pattern; then use this result as input, combined with data conversion strategy SDMS automatically generated a complete evolution from the original model to the target model data conversion steps.3 This evolution of the model as an application upgrade to deal with a long-term process, the evolution from time to time to set up a point, the data conversion step evolution of points assigned to all of them, to get a global optimal consumption pattern of gradual evolution strategy; relational database for the unique complexity of the data conversion using SQL statement forms, and through the example of TPCW standard to expand the idea of the article after the experiments carried out to obtain good results.