Data base concurrency control: versatile approaches to improve performance
This dissertation deals with new approaches to concurrency control that improve system throughput, and discusses some performance issues in data base systems. Firstly, the concurrency control problem is introduced, and some well-known solutions illustrated. The disadvantages of standard approaches to the problem are stated and some newer techniques outlined. Two new concurrency control protocols are described, the Functional Update protocol and the Five Color protocol. The Functional Update protocol is a technique that can be used in timestamp-based systems to reduce abort rates and hence improve performance. This protocol uses semantic information about updates performed by transactions to achieve low abort performance. The Five Color protocol is an unconventional concurrency control protocol. The locking strategy in this protocol is non two phase, and it can be used on any general (unstructured) data base. Non two phase locking was restricted to structured databases, and was thought not to be possible on general databases. However, this protocol illustrates how information about data to be accessed by a transaction can be used, to improve locking performance, and to allow non two phase locking.
- Research Organization:
- State Univ. of New York, Stony Brook (USA)
- OSTI ID:
- 6170744
- Country of Publication:
- United States
- Language:
- English
Similar Records
Performance modeling of data base management systems
Performance measurement, modeling, and evaluation of integrated concurrency control and recovery algorithms in distributed data base systems