Database management systems (DBMSs) are playing an important role in the design of information systems and other applications. A multiuser environment is becoming a standard requirement for applications with the spread of local area networks. Concurrency in such multiuser situations is an important issue that is being analyzed by a number of researchers.
This paper is an important contribution to research in concurrency. The authors propose solutions to update of a transaction at the conceptual level, making use of semantic information implied by a specific database.
The concept of serialization of transactions is considered under concurrency. A schedule for serialization of a set of transactions has the same effect as the execution of the transactions in some serial order. Such problems require exponential time (are NP-complete). The authors consider “domain-based” transactions, where the values of an attribute are independent of the values of other attributes of the tuple. Suitable theorems are stated for serializable transactions. The authors further develop different levels of serialization, such as local and intermediate serialization.
Concurrency control can be further improved by using functional dependencies (fds) that exist in a database. The authors analyze the concept of serialization for dynamic scheduling, in which transactions that are going to operate concurrently are not known in advance to the scheduler. An algorithm for such a situation is proposed. A number of further extensions, including dynamic scheduling using fds, are possible with the concepts from the present research.
Such attempts to increase concurrency have the effect of increasing overhead in processing. Further cost/benefit analysis considering transactions and data is necessary. This analysis is indicated as a topic for future research.
Graduate students and researchers in concurrency will find this paper useful. The list of references is valuable. Such original ideas will definitely make the use of DBMSs more attractive in future applications.
