A recovery scheme for multidatabase systems
CIKM '93 Proceedings of the second international conference on Information and knowledge management
| Hi-index | 0.00 |
This paper deals with the problem of ensuring correct (i.e., atomic, serializable, and durable) transaction executions in a distributed multidatabase system where few changes are permissible in the constituent local database systems to accommodate the demands of the distributed environment. The degree of control over the transactions by the sites (execution autonomy) and by the multidatabase software (control autonomy) are examined to highlight the trade-offs necessary to obtain correct executions. We propose techniques that infringe upon control autonomy in order to provide fault-tolerant transaction management without restricting the types of transactions allowed, and which need minimal changes to be made to the existing systems. These techniques allow a large number of concurrency control protocols to be handled, and local execution autonomy to be preserved. Our proposed protocols tolerate failures at a level comparable to traditional distributed database management systems. Moreover, our scheme exhibits the desirable properties of avoiding global deadlocks, and scalability.