Research Of Data Eventually Consistent Strategy Based On NoSQL

The increasing amount of data and its demand on accessing speeds, especially with therapid development of SNS and Web2.0, makes the traditional database encountering seriousbottleneck problem when users read or write data from the system. The traditional solutionis the distributed technology, who allocates data to different nodes by distributed algorithm.Due to the ease of management and replication, Master-Slave architecture is generally usedby distributed cluster. Because replication in this architecture is a passively asynchronousreplication, it causes a certain time delay problem, and resuts in incoincidences betweenmaster data and slave one. Based on the relational database, this paper introduced a NoSQLdatabase, a caching-database who is on the basis of Key-Value model, to make reading andwriting data more effective. The consistency of the data is a important standard fordistributed database, and the consistency study on NoSQL database model is a new areawith researchers’ serious concern. Inspirationally, success has been achieved in some specialareas, and research in the strategy of data eventually consistent based on NoSQL has veryhighly academic significance.This paper presented an rapid-developed NoSQL database modal based on therelational database, and constructed a system architecture combined NoSQL and relationaldatabases, where memcached was used as a caching layer for MySQL. This paper address amethod to slove the problem of data-replication delay, which used the control of clientaccess of memcached and MySQL triggers Strategy. Under this architecture,when systemhave request of select, it executed commands in memecached and returned to the client atfirst,while the result is not in memcached, it went on executing commands into theunderlying database and wrote into memcached;if the request is modified commands, itexecuted commands in the primary database because of control of memcached client,thenthe tigger based on row record make use of the programmed udf keeping a consistent withthe primary database. The method of paper managed consistency problem of data using theMySQL UDF and triggers, due to the flexibility of UDF and the efficiency of triggers, so itis very efficient method to achieve the underlying database and cache database dataconsistency problems, which can alleviate data inconsistencies problem caused by datareplication in master-slave architecture. Client programming complexity will decline by tigger strategy and memcached eliminated objects by the use of LRU mechanism, it ensurea higher hit rate.Through experiments data analysis, the design has good performance, it can be verytimely to ensure data consistency requirements, and can simplify client programmingcomplexity.