The Research And Implementation Of Distributed Data Replication Over Slow Network Connections

With the rapid development of enterprise informationization an enterprise's information management system should be deployed in the quite wide and scattered spots. A distributed database system takes the necessary priority. In the distributed database system, in order to improve the efficiency and reliability as well as to alleviate the network burden, the data replication technique is adopted so that specific database duplications are saved on every node.Currently due to the various condition limits a slow network is still an important adoption as the enterprise applicable network and remote network. The existed distributed database replication techniques don't take the slow network's characters such as bandwidth and reliability for fully considered, so they are not applicable for the slow network scenario. Furthermore, the existing replication schemes can neither satisfy the enterprises' flexible replication requirements nor support the replication between heterogeneous databases.This thesis proposes a new data replication mechanism, which can solve the bi-directional distributed data replication over a slow network connection. Having completely considered the slow network characters, this scheme adopts the solutions including "single-center-node/multi-replication-nodes" mode, timestamp change capture and two-step update policy. Its advantages are applicable for many kinds of database vendors, little data transportation cost, and low network traffic and highly robust to less reliable network. It can also be used to customize the replication rules flexibly according to the users' requirements. This solution has been applied to an enterprise's database system called "Remote Marketing Management System of Inner Mongolia Sanlian Chemical Co. Ltd." under the low network scenario and worked well.