Asynchronous data translation framework for converting relational tables to document stores

摘要

Although Not Only SQL (NoSQL) techniques have some new features such as query optimization and caching in the era of big data, it is not mature enough to replace traditional Relational Database Management System (RDBMS). In a long period of time, the hybrid solution of RDBMS and NoSQL is a main steam technology. In this paper, we present an asynchronous data framework to translate relational tables into schema-free document stores. The challenge of this framework is how to ensure that the translation is transparent and pervasive to the application with minimal impact on RDBMS performance. To better describe the translation process, we present the intermediate cell state model (CSM) to describe versioning incremental state. We address this challenges by translating the binary log of RDBMS to CSM repository, and the repository is reset to the NoSQL when encountering schema adjustment asynchronously. Finally, we leverage the CSM rectification algorithm and NoSQL synchronization algorithm to implement the whole framework. The experimental results illustrate that this translation has a less minimal impact on RDBMS performance and lower cost than other solutions.