A Middleware for Transparent Access to Multiple Spatial Object Databases

Sang K. Cha, Kihong Kim, Changbin Song, Jookwan Kim, Jooyong Jun and Yongsik Kwon
Knowledge and Data Engineering Lab.
School of Electrical Engineering
Seoul National University
San 56-1 Shinrim-dong Kwanak-ku Seoul 151-742 KOREA
E.Mail: chask@kdb.snu.ac.kr

Extended Abstract

The need for accessing multiple datab ases arises frequently in the geograph ic information processing domain becau se a single database may not contain a ll the desired information in the righ t level of detail, precision, and corr ectness. For example, in planning the construction of underground utilities such as gas and electric power lines, it is necessary to access databases of existing and planned utility networks , and such databases are usually maint ained independently by the individual companies. It is also common that th ese databases maintain much of geograp hic information redundantly with diffe rent levels of abstraction, completene ss, precision, and correctness. Many important decision making processes ca n take advantage of such redundancy by combining the content of one database with those of others to match the des ired level of detail, completeness, pr ecision, and correctness.

This paper describes an on-going effo rt of developing SDBC(Spatial DataBase Connectivity) as a middleware for sup porting multiple spatial database acce ss in the client/server environment. A single application programming mode l based on the spatial extension of th e ODMG object database standard provid es application clients with transparen t access to multiple spatial object-or iented database management systems(OOD BMS). To free the programmers from t he burden of knowing the details of in dividual OODBMS implementation archite ctures, SDBC also defines a server lay er that encapsulates the difference an d commonality among OODBMSs. In addi tion, the global transaction managemen t scheme of SDBC ensures the consisten cy of SDBC transactions accessing mult iple databases. Finally, a framework for spatial integrity constraint defi nition and checking is being developed as a part of SDBC to ensure the quali ty of spatial databases. Sponsored by the Korean government as a project of the NGIS(National GIS) D atabase tool development initiative, S DBC has been under development since t he December of 1996 together with SDBX (Spatial DB eXtension), which provides a set of spatial object classes, oper ators, indexing and clustering schemes on top of commercial OODBMSs. Two o ther projects under this initiative ar e devoted to the development of a dedi cated spatial OODBMS. Figure 1 shows the relationship among SDBC, SDBX, and the spatial OODBMS being developed. It also shows how an existing commerci al GIS system can be incorporated on t op of SDBC.

Figure 2 shows the layered SDBC archi tecture in the current implementation. It includes the OMG Object Request B roker(ORB) layer for the transparent n etwork-level access. Thus SDBC server s are CORBA object implementations reg istered in the ORB, and the ORB direct s the client requests to these servers . However, the client application pr ogrammer does not have to know the exi stence of ORB. The SDBC client API hides the ORB layer. In addition to SDBC servers, the current SDBC impleme ntation also includes the SDBC coordin ator for maintaining the global schema information and coordinating the glob al transaction management.

Although the name SDBC reads similar to ODBC and JDBC, it differs significa ntly from the latter two in the level of abstraction provided to the applica tion programmer. In summary, it prov ides the user with an object-oriented view of spatial databases, global tran saction management, and integrity cons traint checking. At the same time, it shields the user from the burden of k nowing the details of target OODBMS ar chitectural details.

An SDBC prototype is currently being implemented at Seoul National Universi ty in the SUN UltraSPARC environment w ith C++, ORBIX, and SDBX on top of Obj ectivity/DB. A demo application util izing multiple spatial object database s for finding the optimal vehicle navi gation path is also planned to prove t he utility of SDBC.