【NTT】 Database Management System (DBMS) of DIPS

Over a 20-year period starting with its initial research on databases in 1972, Nippon Telegraph and Telephone Public Corporation (now called Nippon Telegraph and Telephone Corporation (NTT)) worked on research and practical application of database management systems (DBMSs) that could promptly respond to the needs of the respective times. A total of 130 users employed DIPS-related DBMSs (approximately 90% of DIPS users employed those DBMSs) on approximately 1,000 CPUs, and the total database capacity exceeded approximately 5 terabytes.

(1)Start of research and development of database management system
In 1972, research on DBMSs was started as part of the DIPS application program research.
 In 1973, the Electronic Communication Laboratories (ECL) of NTT started research on the general-purpose search system DORIS as one of the DEMOS-E libraries. Research and prototyping of DORIS-1 were conducted first. Development of DORIS-2, aimed for introduction to commercial service systems, began in 1975. From 1979 onward, it was applied to such systems as the Nikkei News Search system and the CAPTAIN system.

(2)Research and development of network-data-model type DBMS
In 1974, the ECL of NTT started the development work for prototyping a network-data-model type DBMS, “DEIMS-1M,” based on the CODASYL specifications, which were becoming the mainstay of databases worldwide. While DORIS was a kind of DBMS suitable for information provision services based on reference processing, DEIMS was intended to be a full-scale DBMS that would enable addition (insertion) and updating of data online in addition to referencing of data.
 In 1975, development of the DEIMS-2 for commercial service use was initiated, and the system was released to the data communication division of NTT in 1977. Two types of DEIMS-2 were developed: the function-oriented network-data-model type and the performance-intensive tree-structure data-model type with some network-data-model type functions. In January 1979, the Transportation Ministry’s car registration management system came into service using the tree-structure type. Subsequently, DEIMS-2 was introduced into a number of other systems, including a medical information system (network-data-model type) and Chiba Bank’s system (tree-structure data-model type).
 As the effectiveness of the deployment of the DBMSs to the systems gained recognition and the scope of application was expanded to larger systems, demands for DBMSs to deliver improved performance and offer distributed-processing functions were increasing. To respond to such requirements, DEIMS-3, which was a large-scale distributed-network data-model type DBMS equipped with communication functions including distributed database access and synchronous update processing, was developed. In 1983, DEIMS-3 was deployed into the Social Insurance Agency’s system. This system was a distributed database at the world’s largest scale at that time and had a capacity of approximately 400 GB. Subsequently, DEIMS-3 was adopted into a number of systems in NTT, such as the subscriber billing system, the telephone number reference guide system, and the transmission line network maintenance system.

(3)Research and development of relational DBMS
At the beginning of the 1980s, the need arose for relational DBMSs that were applicable to non-routine tasks systems such as OA and decision-making support systems, in addition to routine tasks, which were conventionally supported by the network-data-model type and tree-structure data-model type DBMS. In 1982, development of the DEIMS-4, a relational DBMS, was started. DEIMS-4 was used in a number of systems, including a medical information system that came into service in 1984, and played a major role in the widespread use of the DBMS.

(4) Supporting diverse data types
Around 1980, the application of databases in the field of CAD/CAM, whereby graphic information is processed in addition to numbers and characters, became of interest for database developers in Japan. DBMS development was decided as part of the DIPS project to achieve a multimedia DBMS that could consistently manage multimedia data such as graphics, images, and sounds in addition to conventionally supported data such as numbers and characters, and would be capable of easily handling complexly structured data. In 1983, development of DEIMS-5, which had a multimedia function based on DEIMS-4, started. The first deployment of DEIMS-5 was in a telephone line design and management system in 1985 and an information storage/provision system (ISS) in 1987. Subsequently, the DEIMS-5 was improved many times in order to meet the changing requirements of the market where it was used. In 1988, a database machine, RINDA, was installed in DEIMS-5. RINDA had been developed in parallel with DEIMS-5 in order to speed up processing in which it was difficult to use indexing, which was the scheme of high-speed data access for relational DBMSs. In addition, functions for online reorganization and reconstruction of databases were added in order to achieve 24-hour operation of databases in 1991. Through these improvements, DEIMS-5 developed into a DBMS with a broad range of applications in terms of both function and performance.