Databases Reference
In-Depth Information
15
Issues on Parallel Processing of
Object Databases
Akifumi Makinouchi
Kyushu University
Tatsuo Tsuji
Fukui University
Hirofumi Amano
Kyushu University
Kunihiko Kaneko
Kyushu University
ABSTRACT
Traditional relational database systems are not adequate for storing, retrieving, and
processing the data found in non-traditional database applications. Such data includes
data for visualization, spatial data, and multimedia data, which is often used in World
Wide Web pages. A number of applications requiring such structurally complex data
are not only data-intensive but also cpu-intensive. Object Database Management
Systems (ODBMS) are efficient to handle such data due to an abundance of data
structures, rich functionality for data processing, and the ability to integrate
programming language and databases. Issues involved in distributed and parallel
processing of object queries and parallel object programming language are presented.
An implementation of ODBMS based on ODMG3.0 standard is introduced also.
First, parallel processing of path expressions is discussed. Path expressions are often
used in ODB applications. Second, parallel processing using a proposed language is
discussed, and methods of performance evaluation are presented. Finally, an ODBMS
based on the ODMG3.0 standard with distributed and parallel processing ability is
introduced. The system, named ShusseUo, consists of three components: WAKASHI,
for distributed data storage; INADA, for ODMG3.0-compliant object handling; and
WARASA, for distributed and parallel OQL compiler. The system architecture and
functionality as well as a performance evaluation are presented.
231
