Study Of Erlang-based Key-Value Database

Traditional relational database faced the performance bottlenecks when data and the number of users are explosive growing , mainly due to the complex relationship between the data association and model coupling, resulting in difficult to extend the database, so it cannot take full advantage of the distributed model for improving performance. The Key-Value database gave up the relationship used in the relational database model, by removing the coupling between the data to make the database more suitable with framework for modern high-performance services to achieve high scalability, high performance. However, because of the current limitations and Key-Value may itself be a large number of clusters, Key-Value Cluster in the data center results high cost, relatively low efficiency, which greatly limits the Key-Value application of the cluster, so to enhance computational efficiency and deployment of the cluster density for the industrial sector has an important significance.In this paper, the Key-Value model's core features of the database, application mode and the current computing environment, the efficiency of cluster computing and propose solutions to the problem lies in the deployment density of the heterogeneous platform support. As the popular stand-alone server architecture to focus so much performance, performance / energy than other poor performance led to the deployment of its high cost of data centers, so efficiency must be a good alternative to using the framework to solve these key issues. However, support for heterogeneous platform features also the use of technology choices, system architecture design and development of the techniques put forward higher requirements, such as: the existing software stack support for the new structure is not enough, need to be a lot of development studies ; heterogeneous cluster performance difference greater need to ensure good load balancing cluster of stable operation. Therefore, the existing Key-Value systems are the lack of good support for heterogeneous platforms. The system uses a customized Linux operating system and the Erlang run-time as infrastructure, full use of both open source, cross platform, high stability, high maturity and other advantages, build a stable foundation to run the database; thorough study the Erlang functional programming approach, process models, message passing language features and fault tolerance mechanisms, in order to solve the communication of distributed clusters, fault tolerance and other difficulties; performance difference for large heterogeneous system characteristics, application of the stratification system , hash and other ideological consistency of the overall framework designed to ensure service scalability.Relative to the relational database, the Key-Value of linear expansion of the database can easily get good performance increase; relative to the existing Key-Value database, the proposed design of the database cluster higher scalability, and full use of existing and future heterogeneous machine resources, to better adapt to the data center environment, providing high performance / power consumption ratio, suitable for high-density deployments.