A global-view-oriented approach to directory management in distributed spatial database

Abstract

Based on discussion on general methods of constructing global directory in DSMDS (Distributed Spatial Database Management System), a new approach based on relation model was brought forward. It specifies three main components of the global directory, including directory organization, directory update and extern interface. Then the authors implemented an actual global directory named SDirectoryService as well, key approaches and techniques in implementation were discussed in detail. Some experiments were designed and performed to check the efficiency, correctness and feasibility of the global directory, which reveal the directory works well on each aspect. By using the directory effective global query processing was achieved in DSMDS.

Architecture and key technologies of grid geographic information system

“…Global resource directory is the fundamental component for achieving corporative geospatial computing. We designed and then accomplished a relation model based global directory organization method, which offers an effective way for geospatial resource directory data management, manipulation and update [17] . At the same time, we brought forward a novel index method for finding geospatial computation resource fast [11] .…”

“…At the same time, we brought forward a novel index method for finding geospatial computation resource fast [11] . Furthermore, we designed the methods to achieve consistency of the global resource directory [17] (Figure 3). 3.3.2 Global query parsing.…”

Fully distributed R-tree for efficient range query dissemination in peer-to-peer spatial data grid

“…We use the Two Step Update Mechanism in [1] to deal with the synchronization of FDR-Tree as node join/leave. When node join occurs, the join node would transfer its root MBR to other nodes, the Virtual Global R-Tree data on the nearest nodes would be transferred to the join node at the same time; then the join node and other nodes would merge their existing FDR-tree with new coming data in parallel.…”

“…Typical representatives include the "disk module allocation" method and its variants [7].The objective in all these methods is to maximize the parallelism for partial match or range queries. In our previous work [1], the data distribution is based on grid partition and each grid cell is called a fragment. Each node contains a set of fragment copies.…”