Font Size: a A A

Design And Implementation Of Online Data Replication System

Posted on:2006-02-01Degree:MasterType:Thesis
Country:ChinaCandidate:M Z WangFull Text:PDF
GTID:2168360152982164Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Efficient data replication system is the key part of disaster tolerant system with the least degradation on system performance. By the in-depth study and comparison about existing data replication systems and relevant technologies, we design and implement an online replicating system called LVR (Linux Volume Replicator). With the principle of data consistency, LVR lowers the response time of replication process by optimizing I/O flow on primary site and that on secondary sites. LVR also provides fast data recovery function which help to fulfill service transition and service continuity.This thesis introduces the related research background, explains the basic concepts and designing principles in LVR system, and figures out the structures and work flows. The data processing flow, system management and exception handles about the system are detailedly shown. Then the experiment results are illustrated. At last the model of hierarchical replicating structure is discussed to overcome the drawbacks in LVR, such as critical elements and weak scalability.The main contribution in the thesis can be summarized as follows: (1) Seamlessly embedding into Linux of the replication module. Accompanied with mechanism of virtual device driver in Linux, and assisted by Logical Volume Manager, LVR intercepts the writing requirements and fulfills the mission of replication. (2) Modularized design. Based on data flows of requests, the replication process is divided into several individual parts. This division minimizes the possibility of visiting critical resources and makes the architecture of the whole system more clear. (3) Dual buffer lists. By using two buffer lists, LVR avoids the status of sleeping while holding locks when kernel handles I/O requests. As well this technology can promote the performance for the improvement of I/O parallelizability. (4) Hierarchical replication structure. Treating the topology of the system as a graph and organizing the nodes as the minimum spanning tree, the hierarchical model can propagate data from the root to all the leaves while distributing the load of the original node into the whole system.
Keywords/Search Tags:Data Replication System, Logical Volume, Logical Volume Replicator, Linux Device Driver, Functional Module, Working Mode, Replication Mode, High-Availability, Hierarchical Structure, Radiating Replication
PDF Full Text Request
Related items