SECTION LINKS
Papers Menu
Land on the Web
Network Data by Design
Georelational Architecture
Customer Facing Spatial Systems
Enterprise Spatial Systems
Outside Plant
Spatial in Corporate IT
Workflow Design

Enterprise Spatial Information Systems

Saul Caganoff

Presented at the annual general meeting of AM/FM Australia-New Zealand, 1997

A major trend in the spatial information industry is the stratification and subsequent mainstream availability of previously monolithic spatial application environments. This promises better opportunities for integration of spatial information systems into mainstream information systems. This paper discusses the technological drivers underlying the evolution of Enterprise information technology and shows how the application of this technology to spatial information is revolutionising GIS and driving the business case for spatial information in the Enterprise. It also presents some strategies for taking advantage of the new technology.

With the opening of global markets and increased deregulation, businesses everywhere are feeling the heat of competition. Many are attempting to face this increased competition by radical restructuring of the business process. Experience has shown that restructuring is not a single panacaea but a continuous process involving constant revision of products, partners, processes, suppliers and services. Many successful businesses also realise that a sustained competitive advantage does not follow from being the best at one thing (eg low cost, better quality, customer service), but derives from a synergy of the Enterprises various capabilities[5]. This synergy requires that different elements within the Enterprise operate together more closely than they may have in the past. In this climate, an Enterprise Information System strategy must be flexible enough to keep up with rapidly changing business requirements, and must enable the Enterprise to derive a deep understanding of its business operations, customers and the marketplace. As spatial information systems become more open, businesses need to make better use of spatial information as part of their Enterprise IT strategy. This need is found in businesses who currently use spatial information systems (eg telcos and utilities), as well as other businesses who are relatively new to spatial information (eg banks). Businesses are looking to spatial information systems to facilitate:

better customer support, to enhance existing services and to identify new customers,
dispatch of mobile units for more efficient maintenance,
future planning by correlating infrastructure growth and demographic data,
monitoring of existing infrastructure,
rapid provision of up-to-date information for disaster recovery,
analyse markets for competitive positioning,
provide an underlying visualisation metaphor for any information systems with spatial attributes (eg a marketing or management information system).

Enterprise Information Technology

Faced with the requirement to increase productivity by lowering overheads and raising efficiency, many organisations are pursuing IT policies which distribute information resources across the whole Enterprise. These Enterprise IT strategies are aimed at meeting the needs of new organisational structures, empowering employees, creating better communications channels and eliminating islands of duplicate information and incompatible technology [4].

But implemention of an Enterprise IT strategy is not an easy task. There are a multitude of problems, both technical and cultural. Often the technological problems are the easiest to tackle. Chief among the technological problems, is the openness required for disparate applications in different departments to interoperate effectively. In the spatial arena this is a major problem which has dogged development of the whole spatial information industry[8]. Beyond openness and interoperability, the manner in which applications are distributed around the Enterprise poses a challenging problem for IT architects. What is the optimal partitioning of those applications? How do applications share data structures and how do applications locate necessary services in a consistent manner?

At the implementation level, there is often a long development cycle for Enterprise-wide applications. The development process must tackle technological as well as business issues. There is a complex round of requirements gathering, design, development, testing and end-user acceptance. Projects fail due to a break-down in any one of these phases. To understand how spatial information will integrate with Enterprise IT, we need to understand how these Enterprise level problems are being addressed by new technology. We will then see that the adoption of this new technology by spatial information systems, opens the way for Enterprise level spatial information systems.

Client/Server

Two-tier client/server architectures are the backbone of most Enterprise IT installations. Their benefits are well documented—as are their drawbacks. Early hopes for scalability of client/server systems have been hampered by problems with application partitioning—the data resides in the server database, the display engine is on the client machine, but where does the business logic go? If the business logic resides on the client (a "fat client") then scalability is hampered by system administration costs in supporting a fat client. System management is a costly and difficult problem as variations in hardware architecture, software applications and application versions must be managed across the Enterprise. Even if a uniform environment can be achieved, the problems of synchronising updates across hundreds or thousands of client machines can be considerable.

A prime example of the problems with fat client systems occurs in the spatial arena. Although some GIS systems support client/server architectures, all the application logic resides on the client. The vast majority of decision makers within an Enterprise cannot have access to spatial data because of the high cost of traditional GIS desktop applications. But, if the business logic resides on the server (a "thin client"), then scalability ultimately runs into server limitations. Furthermore, business logic residing in a relational database (eg using triggers and constraints) does not achieve the seperation between data and logic required for Enterprise applications. In a large Enterprise, business rules will vary from one department to another, and may even vary over the life of a project. An example of this variation in business logic can be found in design of outside plant for telcos, where the early stages of a job require less stringent rules on network connectivity than later stages. Three-tier client/server architectures where application logic is seperate from both client and server has been proposed as a solution to these problems. Finally there is the problem of support for complex data types. Many existing client/server environments handle little more than text and numbers. There are no standard mechanisms for raster, sound and video, let alone more structured data types such as spatial data.

IntraNets

Many organisations are standardising on Intranets for their internal Enterprise IT infrastructure because it coexists with their current systems and solves some of the problems inherent in existing client/server architectures. Standards are developing for a variety of data types such as raster, telephony, video. Unfortunately there is still no standard for spatial data. Intranets are server-centric whereby all data and application logic is held on the server and downloaded at "run-time" onto the client. This addresses system management problems. Intranets are also "thin client," requiring a simple browser application at the client end to deliver all functionality. Whether "thin client" is an inherent attribute of Intranets, or a reflection of the relatively simple applications currently possible, remains to be seen.

Although Intranets address these client/server problems, they bring further problems of their own. Server response and scalability is still a problem. There is increased need for network bandwidth, particularly with the advent of high volume multi-media information. Intranets have no inherent security or prioritisation mechanisms. So mission critical network traffic must compete on an equal basis with "browser" traffic. Site maintenance may become a problem in the Internet world where the demand for rapidly changing content can be expensive. Objects

The cost and implementation problems of todays complex applications is being addressed by object oriented development methodologies which use specific features in object oriented languages such as C++, Smalltalk and Java. Objects aim to reduce development and maintenance cost, and increase reuse through better application architecture and modularisation. Object-oriented development methodologies also aim for the ultimate acceptance of software projects by encouraging involvement of domain experts and end-users in the analysis and design process.

But OO remains a scarce and high-cost skill. It is even more scarce in spatial information systems where there is still a high reliance on proprietary languages. Few mission critical applications have been implemented with objects because of the inherent difficulty in extending these applications to a multi-user OLTP environment. This is largely because of the "impedance mismatch" between objects and relational databases, and the lack of scalability, or mission critical reliability of the current generation of object oriented databases[6]. Object Relational Database Management Systems

Object relational databases are a marriage of object technology and relational database technology. These products allow storage of complex data types along with methods which act on those data types, directly in relational database tables. Although the support for object oriented languages and relationships between objects falls short of those available from OODBMS's, these products promise to provide the scalability of RDBMS's for object oriented applications. In addition, there is better support for legacy data in RDBMS's than currently available via ODBMS's. Oracle and Informix have recently released ORDBMS products, but it is interesting to note that many of the ORDBMS concepts were pioneered in specialised spatial-relational database products such as SDM [1].

Communications Architectures - CORBA and OLE/COM

Distributed computing environments facilitate the partitioning and interoperability of applications across an Enterprise. Existing standards include RPC, DCE and CORBA. Microsoft's proprietary OLE/COM environment is also in wide use.

CORBA defines a common transport mechanism for objects within the Enterprise, along with a common set of services which address object naming, security and persistence, among other things. These are aimed at solving the problems of how distributed object-oriented applications can be partitioned, how they can locate services and how they can interoperate. The Open GIS Foundation recently received submissions from members on how CORBA can be utilised for open and transparent distribution of spatial data. The general thrust of these submissions has been to utilise many of the existing low-level facilities of CORBA, plus define a standard layer of "spatial business objects." The ultimate aim is that all spatial applications, which conform to the OGIS/CORBA specification will be able to seamlessly interoperate[7]. Beyond the spatial domain, the Object Management Group is defining a number of industry specific Business Object standards. The conjunction of standard spatial objects with business objects should allow spatial and non-spatial applications to be truly seamless. A strong competitor to CORBA, with a wider market penetration, is Microsoft's OLE/COM which is the dominant mechanism for desktop application integration. Various self-contained objects (components) may interoperate seamlessly within the Microsoft Windows environment. Although OLE/COM has a firm hold on desktop integration, many industry analysts believe it will be 1 or 2 years before the distributed form, DCOM, can provide a high level of integration across the Enterprise [9].

Oracle, Sybase and Informix have all recently announced middleware products or development environments which are aimed at integrating their respective database servers into an Enterprise internet or intranet infrastructure. Oracle's Network Computing Architecture is based on the CORBA standard. Sybase' Jaguar CTS and Informix's New Era support both CORBA and OLE/COM objects. Given the existing RDBMS penetration in the IT industry, these products may prove to be the most accessible mechanism for implementing a CORBA or OLE/DCOM Enterprise IT infrastructure for many businesses.

Java

Java, the object oriented development language from Sun, addresses problems with hardware incompatabilities by running code within a virtual machine resident on each computer on a network. Thus platform independant applications may be developed for the Enterprise. Java also supports server-centric adminstration of a network by allowing clients to automatically retrieve the latest version of an application when it is invoked [10]. Java and CORBA appear to be a natural match within the quest for Enterprise IT and virtually all major CORBA vendors have developed Java bindings for their products. Oracle has placed Java in a central position as a core development language for its Network Computing Architecture.

Trends in Spatial Information Systems

Given these trends in mainstream IT, what are the implications for spatial information technology? Within the next five years spatial information systems will merge with mainstream IT. This comes about because of the opening of spatial information systems to embrace modern IT concepts (such as openness, client/server) and the commensurate ability of mainstream IT to handle higher volume and more complex data types. A special case application of the ORDBMS architecture may be found in a number of spatial relational databases currently on the market. These include Convergent Group A-P's Spatial Data Manager (SDM), ESRI SDE and MapInfo's SpatialWare. All these systems support storage and retrieval of spatial data in a standard relational database, using extensions to standard SQL syntax. In addition, SDM also supports topological relationships between spatial primitives for "mission critical" spatial information systems[3].

Spatial-relational databases are the catalyst for the current stratification of spatial applications. Freed from proprietary file systems, spatial applications are now being developed for client/server architectures. Separation of database from the display engine, gives spatial information systems the ability to become more "componentised" with application logic being a key and seperable "middle tier" within the architecture. Graphic display engines are becoming a commodity item on everyone's desktop (fat or thin client) and spatial information may now be distributed throughout the Enterprise. With the seperation of costly application logic out from the database and display engine, casual users enjoy inexpensive access to a spatial database. Higher powered users with roles ranging from analysis to CAD and maintenance may use their own specialised applications on the same database as everyone else.

At the desktop, integration of spatial and non-spatial applications is now possible using OLE/COM as the underlying integration mechanism. MapInfo, ESRI MapObjects and Intergraph Jupiter all provide spatial components which support the OLE/COM object model. Integration via the CORBA standard is likely to come about as a result of OpenGIS standardisation efforts, and the need for heavy duty integration of spatial and non-spatial applications beyond the desktop.

SDM provides a key example of the power of this stratification of spatial applications when it acts as a spatial-relational back-end supporting a variety of display engines ranging from high-end CAD systems such as MicroStation, down to Java viewers within a Netscape browser. Consider the following scenario for Enterprise spatial information systems. From a single corporate database, the GIS department may maintain data using a CAD application. The marketing department or MIS department may use specialist analysis applications (often written in-house) using the desktop mapping functionality of MapInfo or ESRI MapObjects. Customer service representatives may view an up-to-date database across the Enterprise intranet using a Java applet which allows them read-only access to the database. A key point here is that choice of a desktop application no longer predicates choice of the backend server. For example, SDM will maintain topological integrity, even if the desktop application has no concept of topology.

Also notable is the recent scramble by major GIS vendors to gain a web presence. This has mostly been in the form of raster representations of spatial data within HTML pages. Unfortunately, raster implementations lack the inherent interactivity required for any spatial applications which can only be achieved by distribution of vector data over the Web. There are only a few vector-based spatial data viewers which operate over an intranet and most of these rely on proprietary backend file servers rather than the newer spatial-relational database technology. The SDM viewer provides the best of both worlds. In the spatial marketplace, there will be further commoditization and componentization of spatial applications. This will drive spatial information further into the IT mainstream to the point where the border between spatial and non-spatial applications will be difficult to define. Spatial information systems will become more heavily integrated with mainstream IT applications. Obvious candidates include workflow applications, document management, SCADA and trouble outage systems within the Telco and Utilities industries.

In mainstream IT, intranets and the internet will become more interactive and used for more complex activities than simple information distribution and marketing. A higher level of interactivity, however will only come about when HTML Web pages are replaced with more powerful environments based on Java applications and/or the services which will become available in distributed object environments such as Oracle's Network Computing Architecture. Spatial information will continue to play a strong role within these environments, largely because of the use of spatial location as a familiar visualisation metaphor for accessing and referencing non-spatial data. Examples of the use of spatial data as a visualisation cue are the Telstra Yellow Pages which give map references for business addresses, or SEQEB's Web page which gives public access to outage information.

Widespread use of spatial information within, and outside the Enterprise is, of course, predicated on the wide availability of spatial data. There must be some investment in an infrastructure for spatial data collection and distribution. The United States has embarked on development of a National Spatial Data Infrastructure, recognising spatial data as being of national importance [2]. In Australia, the deregulation of State land information agencies will have the effect of making spatial data more readily available. Geographic Data Victoria is using SDM spatial-relational database technology to warehouse spatial data for all of Victoria with the aim of making the data available across all government departments and to the general public via internal networks and on the World-Wide-Web. With the continuing growth of the Web as a data distribution mechanism and the success of Open GIS standards, we may see the day when businesses need not maintain local copies of a spatial database, but will overlay their spatial applications directly on spatial layers sourced from an online data vendor. Given that spatial data capture and maintenance is one of the most costly efforts in any Enterprise spatial information system, open availability of spatial data will reduce or eliminate one of the key barriers to spatial information systems in the Enterprise.

Strategies

So what strategies need to be in place to take advantage of the next generation of spatial information systems within my Enterprise? First, there must be an Enterprise IT strategy which encompasses all business units.

Next, there must be investment in a spatial data warehouse which can be accessible to all units within the Enterprise using spatial-relational database technology - flat file systems simply will not cope with Enterprise IT. Setting up a spatial data warehouse means converting existing spatial data or sourcing/capturing new spatial data. This is likely to be a costly exercise, and should be done after a careful audit of data requirements and existing data use. The payback can be significant though, as a spatial data audit in a large organisation will often uncover costly duplication in spatial data licensing and maintenance. For the spatial database to be useful to all units within the Enterprise, you must be certain that the database management system is compatible with applications which will be used (current and future). Most spatial-relational databases currently on the market will only communicate with one brand of display engine/application. If you have a variety of different spatial applications already in use in the Enterprise, then you should choose a database which handles them all.

Finally, there is likely to be some reskilling required within the organisations. Previous pockets of GIS expertise in one or two proprietary GIS systems must augment their skills with some of the open technologies which underpin the new spatial information systems. Relational database design and maintenance is a key skill, as will (increasingly) object-oriented analysis and design. Proprietary application languages will gradually be superseded by "industry standard" languages such as C/C++ and VisualBasic. At the desktop, a knowledge of OLE/COM is useful for application integration. And if you have an intranet or internet site, then HTML, Java and Javascript will be useful. In the longer term, Java, CORBA and ActiveX will be important skills for the development of large Enterprise IT applications on the corporate intranet.

Conclusion

A number of technologies ranging from client/server architectures through to distributed objects and intranets are addressing the business need for more sophisticated Enterprise IT implementations. Concurrently, the opening of traditional GIS systems to client/server architectures and object-oriented integration environments is enabling true spatial information systems to be an integral part of the Enterprise. This will have a number of benefits across the range of customer support, facilities management, market analysis, management information, and real-time SCADA systems. The expansion of functionality will be accompanied by consolidation of resources into an Enterprise spatial-relational database, along with all other Enterprise data. This paper has given an overview of the technology drivers, a peek into the future of spatial information systems and discussed the first steps of an IT strategy for taking advantage of the new technology.

References

[1]	Abel, D.J. (1989). "SIRO-DBMS: a database too-kit for geographical information systems." Int. J. Geographic Information Systems, Vol. 3 no. 2, pp103-116.
[2]	Clinton, W.J. (1994). "Coordinating Geographic Data Acquisition And Access: The National Spatial Data Infrastructure", Federal Register, Volume 59, Number 71, pp. 17671-17674
[3]	"SDM White Paper" SPATIALinfo, 1996.
[4]	Darling and Semich (1996). "Wal-Mart's IT Secret: Extreme Integration", Datamation Volume 42, Number 17.
[5]	King, W.R. (1995). "Creating a Strategic Capabilities Architecture", Information Systems Management, 12(1), pp. 71-74.
[6]	"Comparison of Relational and Object Database Management", On the Sunrise Project Web Site (http://www.acl.lanl.gov/sunrise/dbms/index.htm), Los Alamos National Laboratories, first posted June 1995.
[7]	McKee, L. (1996). "OGIS Spans Distributed Computing Platforms", GIS World, Vol. 9 No. 5.
[8]	"The Open GIS Consortium" OGC's Brochure, http://www.ogis.org/brochure.htm (1996).
[9]	Orfali, Harkey and Edwards (1996), The Essential Distributed Objects Survival Guide, John Wiley & Sins, Inc.
[10]	Tribble, B. (1996). "Java Computing in the Enterprise", Sun Microsystems, September 1996.