| Concurrency control is one of the major issues in database management systems. Although many concurrency control schemes have been developed, most of them suffer from the following drawbacks. First, not many concurrency control schemes designed for distributed database environments truly incorporate the parallelism of distributed systems. To utilize the power of a distributed system, not only the processing, but also the control should be distributed. Second, there is lack of a uniform model for describing concurrency control schemes without ambiguity. Only by using a uniform model, can various schemes be specified more clearly.;In order to describe concurrency control schemes in a uniform way, we first propose a formal specification methodology to describe them. The specification method is based on CSP, which is capable of describing the behavior of concurrent processes in distributed systems. To show how concurrency control schemes can be specified using the specification model, both pessimistic and optimistic schemes have been specified using the specification model as examples.;We also propose a concept of distributed control as the design philosophy. 'Distributed control' means that the control is distributed in such a way that both data items and the transactions that access the data items cooperate with each other in order to maintain the consistency of the database system. To demonstrate the effect of distributed control, we have developed two new concurrency control schemes based on it: one for optimistic concurrency control and the other for pessimistic concurrency control. Thus, another contribution of this thesis is the development of two concurrency control schemes that are more suitable for distributed database systems. These two new concurrency control schemes are first described in usual informal fashion, and then specified using the high level specification. To show the advantages of our schemes, simulation results are included for comparison. |
PDF Full Text Request