/ Subhash Bhalla / Associate Professor
Technological advances in the area of data modeling and design of information systems, aim to capture a wide variety of applications into the fold of database systems technology. In order to provide flexibility in terms of data representation, the databases need the support of powerful data-modelling techniques. Most of the advances in techniques concentrate on capturing more meaning within data that is made resident on a computer system. Within this framework, a number of researchers are actively developing improved data management strategies. The database systems laboratory is participating in the program management of an international Conference on Information Systems and Management of Data (CISMOD 95), to be held in 1995.
The laboratory has activity in progress in the area of transaction processing and modelling of data objects for new applications. In case of new applications, different applications have varying needs in terms of data representation and operations that need to be performed. Many applications add many types of transactions and many types of processing constraints.
The transaction processing environments occuring in various applications include, real-time systems, distributed database systems, and applications requiring complex data representations based on Object-Oriented databases. The transaction management issues among various application environments consider successful completion of user transactions. There are a number of transaction processing domains being explored by researchers in this line of activity. These include - long duration transactions, time-critical transactions, committing transactions in the event of failures, and management of replicated databases.
There is a continuing evolution of computer systems as a hierarchy of computing elements that cooperate to perform computations. In addition complex software systems also introduce multiple layers of software. Thus, the activity of performance evaluation of transaction processing systems is an important area that provides improved understanding of systems. This facilitates improved integration and refinement in technique.
The Database Systems Laboratory has research activity concentrating on data modelling as well as transaction processing activity. A brief outline of the ongoing projects is given below.
Data Modelling for Complex Objects
Information Processing Systems of future will be a combination of integrated components. There will be components for intelligent problem solving, or decision making, components for specialized data processing (e.g., image enhancers), and components for shared information management. The applications will utilize a shared base of information. The examples of some of the application domains are business automation, industrial automation, computer-aided design and manufacture, and cartography. Under the "Data Modelling for Complex Objects" the new approaches for representing such objects are being explored. Among these, a relationship between graphics and databases is being studied.
Real - Time Databases
One of the new challenging area of interest in transaction management is that of Real-time database systems in which transactions have time constraint. Within the new environment, the database management system must support a mix of ordinary as well as real-time transactions. The transaction support must also be highly reliable and fast.
Traditional approaches to transaction management introduce elements of unpredictable delays during transaction processing. Thus, making these not suitable for adoption in new application environments. Many new approaches that are more appropriate for the new environment are being studied.
Refereed Proceeding Papers
Concurrency control based on conventional techniques, requires additional efforts for deadlock detection and elimination. The possibility of a deadlock is also connected to introduction of delays, and repeated restarts of transactions in deadlock cycles. In the proposed approach, a technique for generation of data flow precedence graphs among transactions at data sites, has been studied. The local access graph approach is a fully distributed approach. Through local computations, the approach can prevent deadlocks, in a distributed system.
Distributed systems use optimistic message logging for recovery from transient process failures. Such a recovery is facilitated by asynchronous message logging and check-pointing. It is also supported by garbage collection which requires identifying messages in stable storage that are no longer needed for the process of recovery. For this purpose, it is necessary to keep track of message dependencies between process states. A model to keep track of state dependencies using dependency graphs has been proposed in this paper.
Many applications require support of temporal objects within a database system. These also require database operations that are suitable for executing queries on historic data. Some of the operations within a temporal database system can be defined in keeping with the relational model's interpretation of existing relational operations. Based on this notion, a set of fundamental relational operations has been studied for adoption within a three-dimensional temporal data model. Adoption of this proposal, results in a consistent set of operations, and also, eliminates some of the anomalies associated with the earlier proposal for supporting binary join operation in a temporal database management system.