Yasuo Yamane

An object-oriented database system Jasmine: implementation, application, and extension

We have devised an object oriented DBMS called Jasmine for advanced applications. The paper describes the implementation, application and extension of Jasmine in detail. We focus on the impact of the design of its object oriented model and language on database implementation technology. We describe what part of traditional relational database technology we extend to handle object oriented features such as object identifiers, complex objects, class hierarchies, and methods. We introduce nested relations to efficiently store and access clustered complex objects. We use hash based methods to efficiently access nonclustered complex objects. We provide user defined functions directly evaluated on page buffers to efficiently process method invocation. We devise object oriented optimization of queries including class hierarchies, complex objects, and method invocation. We incorporate dedicated object buffering to allow efficient access to objects through object identifiers. We also describe nontrivial applications of Jasmine and discuss the validity of object oriented databases. We focus on a constraint management facility, which can be implemented by taking advantage of the extensibility of Jasmine. The facility includes constraint rules, called design goals, for automatic database population required by engineering applications. Finally, we describe a view facility for schema integration also needed by engineering applications in distributed environments. We focus on how we extend Jasmine to implement the facility.

Parallel Partition Sort for Database Machines

We have been researching database machines based on parallelism using a multi-processor. In this paper, we briefly describe our database machine architecture and then we describe the parallel sort method which we are investigating based on that architecture. We discuss a new parallel sorting method, called a parallel partition sort, which transfers only a small amount of data and does not place large demands on the CPU. This method is based on the top-down partitioning of data. We experimented and evaluated our method using a multiprocessor to compare our method with the block bitonic sort method.

A Hash Join Technique for Relational Database Systems

In this paper, we will formulate and discuss an efficient method of equijoin in relational database systems. In this method, if either of two relations can be loaded into internal memory, the equijoin can be processed quickly using a hashing technique. Otherwise, the relations are recursively partitioned into subrelations using hashing to be loaded. We also analyze the method and evaluate the results of our experiments. We will concentrate on how we should partition relations.

Multi-vector feature space based on pseudo-Euclidean space and oblique basis

The Earth Mover’s Distance (EMD) and the quadratic-form distance (QFD) are representative distances used in similarity searches of images. Although the QFD greatly outperforms the EMD in speed, the EMD outperform the QFD in performance. The EMD, however, has almost no theoretical justification and requires high computation costs. We propose a feature space model we call a “multi-vector feature space based on pseudo-Euclidean space and an oblique basis (MVPO).” In MVPO, an object such as an image is represented by a vector set (roughly speaking, a solid) and the EMD is reinterpreted as the distance between vector sets while the QFD is reinterpreted as the distance between the centroids of vector sets. Therefore MVPO gives a common geometrical view to these distances. We hypothesized that in MVPO the entity of an image is represented by a vector set (solid) and geometrical reasoning is applicable to MVPO. Our hypothesis explains well that the EMD outperforms the QFD in performance because the centroid of a solid is the simplest approximation of it. Our hypothesis implies that the performance of the QFD should be good when solids are far apart but bad when they are close together. We conjectured that discriminability would decline—that is, dissimilar images would be judged to be similar—when the centroids of solids are very close. Our experiment supported this conjecture. And from our hypothesis we conjectured that by making an original solid simpler, we can make an approximation method that has better performance than the QFD and faster than the EMD. The results of our experiment with this method supported our conjecture and consequently our hypothesis.