Design and evaluation of a relational data base machine employing advanced data structures and algorithms

This paper presents the architecture design and evaluation of a relational machine employing advanced data structures and algorithms to manipulate set-level operations required in a relational model. We have developed data structures and algorithms that are suitable for VLSI-oriented parallel processing. The data structure called VTF (valid tuple flags) is a simple bit string that is used to distinguish valid tuples from invalid ones. VTF helps to reduce the cost to recognize the valid tuples, to minimize the memory requirement, and to avoid constructing a temporary relation as a result of executed relational operations. By extending concepts of VTF, we have established an advanced processing method based on efficient data structures and algorithms that could avoid reconstructing relations and numbers of iterative searches required in accomplishing relational operations. This paper describes basic concepts of VTF, an efficient processing method using VTF, design of a relational machine architecture introducing this method, and performance evaluations of the method and the architecture. Basic concepts and extended concepts of the method incorporated with VTF will provide strong foundations on architecture design of a relational machine.