A Study Of NOSQL-oriented Performance Analysis System

With the development of Internet applications, especially in recent years Big Data applications and unstructured data or changing data continue to emerge, which put forward new demands for data storage. Traditional relational database is gradually revealed some defects, and cannot well meet these requirements of new Big Data applications. NOSQL (Not Only SQL) is a new non-relational mode of distributed storage technology, using new data model, with huge amounts of data storage, high-performance, high availability, strong scalability and other features. NOSQL technology is not intended to replace the traditional relational database systems, and the two have different application scenarios. Many NOSQL storage systems have been used for Big Data applications. In recent years, with a growing number of corporates and personal attention on NOSQL storage systems, performance analysis and tuning has also become a hot research focus. As a result, how to effectively conduct performance analysis and evaluation of NOSQL storage systems in the distributed environment in order to achieve performance optimization have a major impact on the development of NOSQL.In this paper, we conduct analysis of existing methods of performance analysis and propose a new NOSQL-oriented performance analysis approach considering the defects of existing methods. In the distributed environment, relating the performance of NOSQL client to data processing on the server side to conduct in-depth performance analysis and evaluation of NOSQL storage systems. Monitoring the runtime behaviors and data processing activities using dataflow-based distributed instrumentation mechanism, and associating server side data with client performance data on the same timeline to conduct in-depth performance analysis. which could assist developer to find and identify potential optimization points. The NOSQL-oriented performance analysis system mainly has three advantages. First, it is lightweight and has good scalability. Second, it has no source code modification and no any modifications to the system runtime, middleware, or messages. Finally, our performance analysis system is low overhead and has negligible performance impact on the running NOSQL system. We report our experience case on how our performance analysis system helps user to efficiently conduct NOSQL performance analysis and tuning.