Fault Detection And A Copy Of The Digital Organism Database Recovery Mechanism And Realization

With the increasing widely application of distributed database in business area, customer's requirement for the reliability and utility of database is becoming higher and higher. Therefore, it's crucial meaning for related service provider to develop a distributed database system with robust fault-tolerant capability.With over years experiences of research and development in distributed and paralleled area,8010 laboratory of UESTC have developed a new type of distributed database system called DOSSQL, which combines with the distributed feature of P2P system. Building on open-source database system ---- MySQL,it not only maintains the high process speed of MySQL, but also attains much new advances such as, high reliability, high data throughput, high storage capabilities.Fault-tolerant mechanism is a guarantee for previous advances of DOSSQL. It includes data-redundancy, failure detection, failure recovering scheme. Among those, failure detection is a base for failure recovering scheme. Failure recovering is the last implementation part of fault-tolerant mechanism. In DOSSQL, failure recovering scheme includes both restoration of system structure and restoration of data on re-joining notes which have been debugged. DOSSQL take data-redundancy strategy to keep multiple duplications for one database, those duplications have an equally status, it therefore would be disaster for whole system, if there is any accordance among those duplications. The purpose of duplications recovering scheme (DRS) exactly is to keep the re-joined duplications, which had been out of system before, in step with other duplications. Therefore, the important role of DRS is self-evident.In this thesis, above on further study on distributed and paralleled database system, we address the Fault-tolerant mechanism of DOSSQL, focus on failure detection, failure recovering scheme. According to the topology and logical organization of DOSSQL, failure detection scheme combines with two kinds of new check technology to detect all notes of system. As for database recovering scheme, after further study on shortages of current covering mechanism, we proposal a new improved covering algorithms. This algorithm combines the database backup mechanism with Log Recovery mechanism, and makes it possible that providing service in best effort for users while database recovering. In this way, it decreases the overhead of log's recording and maintenance effectively.