Deadlock detection and resolution in data-base management systems: A comprehensive approach
In this dissertation, several algorithms for deadlock detection and resolution in database management systems are presented, where two-phase locking -is assumed for ensuring serializability, the lock requests obey the granularity locking protocol and each granule may be locked in one of the following lock modes: IS, IX, S, SIX and X. For each object, lock requests are honored according to a first-come-first-served basis except for lock upgradations. The author presents algorithms for deadlock detection and resolution in sequential transaction processing to achieve the goal of early deadlock detection with the appropriate victim selection. He also presents a deadlock detection and resolution algorithm for parallel transaction processing which achieves the same objectives and an algorithm for distribute transaction processing which minimizes the amount of inter-site message communications. He proposes a new efficient algorithm for deadlock detection in sequential transaction processing, where the basic idea is the construction of a directed graph called a Holder/Walter-Transaction Waited-By Graph. He establishes guidelines for the identification of a victim in a deadlock cycle and show how deadlocks can be resolved with minimal victim cost. In addition, his algorithm allows us to resolve some deadlocks without aborting any transaction. In the case of parallel transaction processing, a transaction can have multiple outstanding lock requests. He introduces two types of deadlocks: explicit deadlocks and implicit deadlocks. To detect deadlocks in this environment, he introduces a new type of directed graph called a transaction waited-by graph. He presents deadlock detection mechanisms, identify deadlock detection time, and show how victims can be selected with minimal cost.
- Research Organization:
- Northwestern Univ., Evanston, IL (USA)
- OSTI ID:
- 5896898
- Country of Publication:
- United States
- Language:
- English
Similar Records
Deadlock avoidance revisited
Deadlock prediction via generalized dependency