ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001
316
A NEW APPROACH FOR DISTRIBUTED GIS
Yuxiang WANG, Chongjun YANG, Donglin LIU
China Beijing 100101
Institute of Remote sense Application, Chinese Academy Sciences
KEYWORDS: Distributed GIS, CORBA, ORB, Geo-Spatial Data
ABSTRACT
The research described in this paper explores a new approach towards distributing GIS tools across multiple CORBA ORBs. The main
structure of a distributed GIS is introduced. How CORBA and GML are integrated is explained. Object migration is implemented among
servers for load balance and security is considered. Then the main procedure of designing a CORBA distributed GIS is introduced.
1. INTRODUCTION
GIS has become an indispensable spatial data processing tool
for government and research. With strengths in reliability,
processing efficiency, resource sharing, parallel processing and
system scalability. A distributed GIS is a collection of GIS sites.
Each site is an autonomous GIS that contains both
geo-processing and spatial data. It is a group of physically
distributed GIS sites. They are logically united as one site. For
user, geographic service can be accessed transparently.
• Compared with a centralized-server GIS, the major
advantages of a distributed GIS are as follows:
• Evenly load balance can be achieved by strong
multiple GIS sites. When a site works wrong, other
sites will finish the process left.
• High speed can be achieved by distributing the whole
job to multiple sites for parallel processing.
• Data and processing interoperability are enhanced and
system development cost is reduced.
• In a well-planed distributed GIS, system is easily
scalable.
A distributed GIS has advantages in data sharing, reliability and
system growth. In recent years, distributed GISs have attracted
increasing interest. For examples, the Distributed Geographic
Information System Project in Australia and in Norway, and the
geodata modeling technique for distributed GISs at Berkley. The
National Center of Geographic Information Analysis(NCGIA)
includes distributed GIS in its research plan. The University
Consortium for Geographic Information Science included
distributed computing and interoperability in its research
priorities.
2. CORBA BRIEF
The Common Object Request Broker Architecture (CORBA) is
made up of Object Request Brokers (ORBS) that can
communicate with each other via the General Inter-ORB
Protocol (GIOP), the Internet Inter-ORB (HOP) Protocol [2].
CORBA facilitates a middle-tier, object-to-object infrastructure
that allows one to encapsulate data from multiple sources. The
language and compiler used to create the server objects, as well
as the location of distributed CORBA objects and the operating
system they execute on, are totally transparent to clients.
CORBA therefore provides an ideal mechanism for creating
3-tier (or n-tier) distributed applications that go beyond providing
simple interoperability [2].
CORBA provides one with ability to perform dynamic discovery
of objects and services, as CORBA objects are self-describing
and introspective. CORBA's dynamic facilities, including
Dynamic Invocation Interface (Dll), and the Interface Repository
allow the creation of extremely flexible systems that allow
run-time discovery and late-binding [2]. This is especially useful
in the Web environment where a user is able to discover new
services and then make use of them transparently.
3. DISTRIBUTED GIS MODEL WITH CORBA
Distributed GIS is based on a distributed geospatial database.
Clients can access GIS server remotely and concurrently. The
whole structure of our distributed GIS is showed (Fig.1).
Because our distributed web GIS servers include many functions,
we divide our GIS functions into different ORBs. They are
projection Transform ORB, Data Format Transform ORB, Spatial
Database Storage ORB, Spatial Index ORB and Application
ORBS (Fig 2). All the CORBA objects are communicated by HOP
protocol. They are located and managed by Naming Service and
Interface Repository. The naming service locates data objects by
name. Each data object, for example, a digital map or a remote
sense image, has a descriptive, recognizable name used for
query. The naming service maps the name into the address and
the reference of the object.
Fig 1 : Distributed GIS Architecher