الفهرس 
AcknowledgmentOnly 14 pages are availabe for public view
 |  | from | 117 |  |  |
| | | from | 117 | | |
Abstract
Perhaps the most neglected objective of a database management system is the maintenance of data quality. Human and system errors are the main reason for poor quality of data. l1aintenance of quality involves correction of the entails precautionary measures and erroneous data. This additional software functions within the database management system in what is called the recovery manager. The major function of the recovery manager is to achieve database consistency despite the exietence of errors [11. Data consistency is violated by several types of errors such as hardware errors (s.g. power failure ), software failures ( e.g. bugs in programs) or deaplocks due to Job interact i etns. Power failure, for example, affects database consistency. Due to the loss of main memory state, complsted jObs before power failure may not have been reflected on the database yet. This may introduce problems in database systems such as banking and flight reservation. It is the job of the recovery manager to recover from such errors. Recovery techniques are used in order to achleve database resiliency. This thesis introduces a method for evaluating recovery techniques in database management systems. By uSlng thlS method for evaluation, analytical equations which model these recovery techniques are introduced. USlng thls model. several performance studies have been presented. have been confirmed by using a eimulator program.
Results In chapter 1, the concept of a transaction as a unlt of recovery has been clarified. Recovery as a transactlon management issue is also surveyed. For the sake of completness, other transaction management issues related to transaction processing in database management systems such as transaction initiation and termination. and concurrency control have also been surveyed brietly. Recovery implementi~tion has been discussed in detall in chapter 2. System R [2] is presented as an example of a recovery manager. Chapter 3 introduces the suggested method for recovery techniques evaluation with the performance indeces proposed. It also discusses previous and related work in the fleld of recovery. As an application of the suggested method. an analytical model of a general database recovery manager is introduced in chapter 4. Several performance studies are performed using this model. and are pres~nted in the same chapter. A simulator program was written for the purpose of confirming results. The simulator outline is included at the end of the chapter.
In chapter 5, analytical and simulation results are contrasted, and justified. Several conclusions are drawn. The suggested method for recovery techniques evaluation as well as L~.e results obtained from applying this method are both criticized in the same chapter.
)