Replication Dynamic Consistency Strategy In Large Scale Data Environment

With the popularity of the internet, lots of internet applications turn into out life, suchas social networking service, online video, photo sharing, and e-commerce, etc., and theWeb2.0time comes. These applications commonly use distributed architecture to storethe large scale data, and how to effectively manage the multi-replications between nodes isa continuous research topic. A dynamic consistency strategy is presented in the large scaledata environment for the reason of consistency requierent to the different applications arediffer from each other, and this strategy can be more flexible and precise description of thebusiness for different consistency requirements.Error metrics exist between multiple copies of replication nodes in thesynchromnization process. Each type of data can be described by one of these three errormetrics: numerical error, order error, and staleness, and a corresponding error value aregiven. The dynamic consistency strategy is ensured by the cooperative work of taskdistribution module, request processing module and scheduling module. The principle ofdynamic consistency is as follows: firstly, the request with a pair metrics is analyzed bythe distribution module and put the task into the priority queue, and then the requestprocessing module takes a task from the queue and executes it by specific requirement.The scheduling module is intended to avoid the lower priority task has no chance to beexecuted.The experiment shows that the dynamic consistency can meet the situation of allerror metrics in one application very well. The dynamic consistency in the large scale dataenvironment can address the different consistency requirements of different applications,but also meet the different consistency requirements in the same application and enanlesclients to control over the consistency more flexible.