Distributed Database Consistency And Availability Methods And Optimization Research

With the advent of the era of Big Data, in the past few years, the world ushered ina revolution in the database, the characteristics of the transaction is not a relationaldatabase in a monopoly position, the birth of a non-relational database can make avariety of network applications are not affected limited in some way, CAP publishedtheory allows us the flexibility to use a different database consistency in theapplication of the policy to ensure the performance of the system.CAP and abroad on the basis of the theory of the birth of many new distributeddatabase, especially non-relational database is sprung up in front of us, so that wehave more choices, and the way these architectures with a database model also gaveus a lot of inspiration.Firstly, the current mainstream distributed cluster strong consistency of therelational database MySQL MySQL Cluster architecture is analyzed and itsperformance test, after discussion of the test results were analyzed to join in makingits Proxy layer (MySQL Proxy, MySQL Amoeba, lifting the lower HiveDB andMySQL Federated) of the system performance; then the domestic IT giant AlibabaGroup R&D OceanBase database were analyzed, discussed in detail its systemarchitecture model as well as its strong consistency in maintenance mode, study Thedatabase system to ensure consistency in the case of data exchange through how toimprove the performance of the system as much as possible, to explore the stability ofsuch a distributed database.Second is HBase, MongoDB and Cassandra three current mainstreamnon-relational database analysis and testing. The main structure of the strongconsistency model for HBase database were analyzed to study how to ensure strongconsistency of its systems and ways to improve system performance, its read andwrite performance and MySQL were compared to test and analyze the test results; tohave efficient reading MongoDB database performance for a simple introduction tostudy the structure and consistency of the way the database strategy, then the read and write performance of the database to do a simple test and analyze the test results;consistency can use a variety of strategies Cassandra database for a brief introduction,and detailed analysis of the consistency of its various strategies and analyze theimpact of the database consistency in case of using different strategies on systemperformance, the last three of these were African relational database comparison test,to discuss their respective application scenarios fit through their own architecture andtest results.Finally, this paper an improved model of a distributed database, the schema ofthe database model is similar to the way OceanBase database, its database consistencymaintenance strategy based on some strategies on Cassandra database consistencymade some improvements, in addition to several increase the consistency of the newstrategy, the database model will be completely separate read and write operations, sothe consistency of different strategies for reading and writing, in addition to strongconsistency of a single node and read operations to write operations outside thesystem and ensure the highest performance read and write In addition, the new logand increase the consistency of a single node while reading level, which is differentfrom the local level in several previous database consistency is the consistency ofchoice to the user, by the user to dynamically select the consistency of policies basedon performance requirements, such you can ensure that users choose good they offerhigher performance consistency level read and write operations under circumstancesrequire, and then use three database cluster simulation model of the database andanalyzed to test.