Design And Implementation Of The Online Physical Full Backup And Restore System For GBase

The development of database with independent intellectual property is an important step for our information industry. There is an urgent requirement of it in national defense and e-government. The GBase DBMS is a homemade database with independent intellectual property, which nanda general company and a graduate school upgrade by their database product archetype. The online physical full backup and restore system that this paper introduces is one of basic peripheral tools of GBase DBMS, which has important meaning on backing up the information system, restoring and popularizing the whole database product.Most database manufacturers have afforded their online physical full backup and restore system. For example, the RMAN of international database Oracle, the online physical full backup and restore tool of homemade database DM. So there is a great requirement to design the online physical full backup and restore system that fits the property of GBase.This system is based on the checkpoint LSN mechanism, which makes it be impossible of integrality of backup and restore. As for online physical full backup, backup starts with initiative checkpoint, and ends with terminative checkpoint. In the process of backing up data files, it implements to backup log record information changed. The content of backup log record is log record information changed between initiative checkpoint LSN and terminative checkpoint LSN. As for restore, restoring is also getting up initiative checkpoint, and terminating at terminative checkpoint, in other words, the process of restoring to analyze the log record begins from the initiative checkpoint LSN, stops at the terminative checkpoint LSN.This system adopts the C/C++ for programming, which has finished the design of Windows and Linux edition. System is made up of two parts: The backup and restore. The backup has mainly four modules: The module of backup analysis, which is used for analyzing the control file, parses data and log record file information; The module of backup data and log is used to backup data file and the log record information having been changed; The module of treating with backup log record file, which is used in arranging backup log record file, realizes the completeness treating of backup log record file; The completeness that backup the table structure files and the control file is the complementarities of backup system. Restore includes mainly five important modules: The module of restore pretreatment realizes the prepared job before actual restore, mainly including clearing up the restore catalog, restoring the data file to allocating path, restoring the log record file to data base path, and restoring table structure files and control file; The module of restoring analysis process implements to analyze log recordand stores them to hash table; The process module of redo is used to apply the hash table data to data page; The process module of undo is used in undoing the uncommitted transaction having been applied to data page; The module of consistency treatment which finishes the last job of restore makes the data file come to an agreement with the log file information.The first characteristic of system is the tactic that the backup data and log record file adopts: the backup of log record must be made once when certain amount backup of data file has been finished. It resolved the problem of data being lost since log record circulation write brings. Because of backup data file is above GB level mostly, in the process of backup, transactions that is in the active session, extremely, have possibility to bring about circulation write of log record file, and then bring about data being lost.The second characteristic of system is the tactic that how process of redo increases data file: to increase data file according to the setting of control file; it is that the size of the data file increased one time must keep according with control file. Such has decreased the rate of the frequency of expanding the data file (if increasing data according to real needing, it will have possibility to bring about this condition that expanding the data file frequently), and does most work for implementing consistency matching of the size that data file records and real size of data file.The third characteristic of system is that making use of the batch treatment command that operation system provides to realize part of backup and restore operation and the job invoking the main procedure. Because of the GBase table structure file and the table space data file are to part, and every table must be corresponding with table structure file, for decreasing complicated degree of programming, implementing backup and restore of table structure and control file using the batch treatment command of OS (operating system utilities) is a reasonable choice, and making use of command of OS has better efficient, stable, flexible characteristic and so on. After completing this backup and restore system, we have carried out detailed science testing on the systematic integrality on windows platform, and have carried out comparison with the same kind of database firm product of homeland and international. The results of test and comparison verified that the system can accomplishes backup and restore of the data successfully on the function, and reach no mistake about testing on million data; the efficiency of the system can achieve and even exceed the level of the native system.There is some work remained to improve and implement, for the system is just the first version and time is restricted. On the function, For realizing the entire backup and restore system, system must get up with organic union of assignment administration, such ability of backup and restore will play outmore well; The backup and restore just has been finished on local, the remote backup and restore will be the next work considered; At present the backup and restore is based on the operation system file rank, which basing on naked equipment hasn't been implemented. On performance, system will finish multithreading fast physical backup. All of the above problems will be achieved in the later version.