الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 230 |  |  |
| | | from | 230 | | |
Abstract
In the recent years, steady improvements in computer hardware and network technology have to lead to dramatic increase in the information technology. Generally, it can be said that sharing information in distributed systems is the most important requirement in the information technology Concurrency control and deadlock detection is the most important problem that must have a powerful attention when sharing information in distributed systems. Several algorithms have been developed for detecting deadlocks in distributed database systems. However, there is no data about the performance of these algorithms.The main objective of this thesis is to address the deadlock detection in distributed database systems, to study the performance of various deadlock detection algorithms, and to introduce new algorithms to detect deadlock with a reduction of the communication overhead through the network during the sharing of data. In this thesis, a simulation model is developed to study the performance of four deadlock detection algorithms. Two of them are selected from centralized algorithms (Two-Phase Deadlock Detection Protocol and One-Phase Deadlock Detection Protocol) and the other two algorithms are selected from distributed algorithms. In this model, the communication overhead and time delay is determined for these algorithms.The thesis introduces the hybrid algorithm. Although this algorithm has more advantages over the other deadlock algorithms, more investigation on it proves that it has some deviations. These deviations are introduced and discussed using examples. The thesis proposes new modifications on the hybrid algorithm. These modifications are proved to be correct. When these modifications are applied, such deviations are eliminated. The thesis also proposes a new deadlock detection algorithm. This algorithm can be classified as distributed algorithm because it doesn’t depend on a central site to detect the deadlock. The communication overhead of the proposed algorithm is introduced in terms of the number of messages, the average length of messages, and the average data transferred and compared with the communication overhead of the chandy-misra-haas algorithm. from the result obtained, the proposed algorithm has smaller number of messages, smaller length of messages, and consequently has smaller number of data transferred than Chandy-Misra-Haas algorithm.