Efficient query result retrieval over the Web

Consider a geographic information system (GIS) which is set up as a Web server that allows users to query the database with a Web browser. As the query result may be huge and the network delay could be significant, we investigate the fundamental problem of how to deliver the query result efficiently over the network. In a conventional client-server database system, the commonly used application programming interface (API) is the so-called iterator-based interface in which a client queries the server with an ISQL statement and the result, which is called a result or active set, is generated. To retrieve the query result, multiple calls are made to the server and objects in the result set are retrieved sequentially. To enhance system performance, objects in a result set can also be retrieved in bulk by storing them in, an array. In the Web environment. a database server is commonly implemented with a distributed object technology such as Java or CORBA. As network delay could be significant and the client memory spaces are limited and varying, neither multiple calls nor bulk retrieval is a viable solution to this problem. We propose a technique by caching and piping the result set through a socket connection without forfeiting the iterator-based interface. We show that the proposed method is superior in delivering a query result in a LAN and in the Web environment. We then investigate how to retrieve and display geometric data in a map efficiently in a network environment. We show by experiments that, compared to other R-tree variants, the class of Multi-scale R-tree is an excellent indexing structure for supporting the display functionality in, a GIS, especially when the system is set up as a Web server.