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

Spatial Information Systems within the Corporate IT Framework

Saul Caganoff, PhD

Presented at the GDS User group Conference, 1996

Abstract

The golden age of the digitiser is over. Users of spatial information now need to make maximum use of their hard-won and valuable spatial data—beyond traditional GIS applications. This sentiment echos general user requirements within the mainstream IT community which has been moving toward open distributed systems for some time. Here, we examine current trends and look at the future of spatial information systems within the Corporate IT framework.

Introduction

The current crisis in spatial information systems (SIS) is well summarised by one of the introductory statements in the Open GIS Consortium (OGC) home page on the WorldWideWeb: "OGC was initially founded...in response to the widespread recognition of the following problematic conditions in the geoprocessing and geographic information community:...[t]he multiplicity of geodata formats and data structures, often proprietary, that prevent interoperability and thus limit commercial opportunity and government effectiveness."

Spatial information systems are yet to penetrate deeply into the corporate framework largely because of the interoperability problems identified by OGC. The reasons behind the multiplicity of data formats and lack of interoperability are many and varied. Part of the problem is that spatial data is complex and difficult to manipulate. The technology required to handle spatial data has taken time to evolve and has followed a number of different paths in its evolution. Today, spatial information systems lie outside the arena of mainstream information technology (IT) which more directly addresses the needs of a distributed enterprise. This paper examines the direction of mainstream IT and argues that this direction points the way for the next generation of spatial information systems.

Where is IT Heading?

Mainstream IT is being driven by user requirements for corporate wide access to resources, interoperability and scalability. Resources here are not only data resources, but development resources—the ability to utilise a development team across multiple projects within the enterprise.

These requirements are being addressed by utilisation of standard operating environments, open systems standards, client server architecture and the use of parallel and distributed processing. Standard operating environments are an attempt by an enterprise to control the constitution of their IT resources. The concept of open systems underpins the development of vendor-neutral computer technology to enable interoperability across hardware platforms and even across software applications.

Enterprise wide access to data requires an architecture that maximises data integrity and ease of maintenance. A centralised architecture satisfies these requirements, but has problems with scalability because eventually the database server cannot handle an increasing load. Client-server architecture has evolved from the old mainframe architecture to increase scalability by putting some of the application or display functionality at the "terminal". This architecture, known as two-tier client-server still has scalability problems in today's enterprises. Technologies to improve scalability include parallel processing hardware to better manage concurrent client connections, and three-tier client-server. The logical development of three-tier client-server architecture is a distributed processing environment where processing tasks are shared throughout the enterprise on an as-needed basis. Examples of distributed processing architectures (future or present) are the Common Object Request Broker Architecture (CORBA), Microsoft's (OLE/COM) object API and message oriented middleware such as Lotus Notes.

Where are Spatial Systems at?

By contrast with the open, scalable, distributed direction of mainstream IT, spatial information systems have a long way to catch up. Spatial information systems (including GIS, AM/FM and LIS) rely on proprietary data structures and languages as well as monolithic architectures.

All major SIS vendors rely on proprietary formatted file systems for storing spatial data. This has a number of major drawbacks for openness and enterprise wide access.

By definition, a proprietary file format is not open. Users must rely on translators between proprietary file formats, some of which cannot handle the capabilities of others. Topology is a primary example of spatial information that is not handled by many of the common translators. The Spatial Data Transfer Standard (SDTS) is an attempt to set up a comprehensive standard for data translation. But this still does not address the proprietary nature of the file storage.

Proprietary file formats also present problems for enterprise wide access to spatial information because of their nature as "flat files." Locking and concurrency present problems for multi-user access and their finite size inhibits the view of the data as a seamless mapbase. An appropriately intelligent file server or data server can be wrapped around the file system to facilitate locking, concurrency and seamlessness—this is essentially a database management system.

In an attempt to alleviate problems with enterprise access to spatial data, most vendors have provided links to non-spatial attribute data within an RDBMS that can be associated with spatial attributes within the proprietary file format. This creates further concurrency problems that can lead to spatial and non-spatial data getting out of step with each other. Backup cycles must be synchronised between the spatial and aspatial data and the already complex task of managing very large systems is made worse.

The second great sin against openness is that most SIS vendors rely on proprietary languages for application development within their product. This increases the learning curve for application developers over and above that required simply to learn the concepts behind dealing with spatial data. Application developers become hard to find and expensive. Knowledge of spatial data and its manipulation becomes a specialist, "boutique" skill.

Spatial information systems are still primarily implemented on a monolithic architecture. The data store (file system), application intelligence and display engine are all on the one box, keyed to one licence and accessible by only one user at a time. This makes sharing of resources (both machine and human) difficult or impossible to the point where large enterprises are excluded from making full use of their spatial information. All these elements reinforce the view that spatial information is somehow "different" to anything else within the corporate IT arena.

Barriers to Assimilation

Spatial information has become ghetto-ised—utilised by only a few small groups within the enterprise. Consequently spatial information is often mis-understood or simply overlooked within the enterprise, by those assembling management information systems, by those carrying out business analysis and by those holding the purse strings. All the issues discussed above must be dealt with in order to bring spatial information out of the back room and into the enterprise IT mainstream. But the main barrier to assimilation is simply the inaccessibility of spatial data and the mindset that inaccessibility creates. There are no other intrinsic impediments to the mainstreaming of spatial information systems as we have already seen with enterprise IT moving from the text-based data processing days into the rather more complex world of multi-media.

Emerging Requirements

This is not new to the users of spatial information systems who have been voicing these concerns for some years now. Their first hand experience of the problems with spatial information systems are driven by increasingly sophisticated requirements for the analysis and manipulation of spatial data. The golden age of the digitiser is over. Capture of spatial data is a problem that is largely solved and the emphasis now moves to maintenance and getting a payback for the investment in spatial data infrastructure.

Examples of the sophisticated analysis required of spatial data are:

visualisation of spatial data which goes beyond the coloured boxes of thematic mapping,
spatial modelling such as predicting the distribution of rare animal species based on vegetation, climate and human habitation,
modelling the connectivity and characteristics of a large telecommunications network.

These operations often use specialist software that correlates spatial data with other forms of non-spatial attributes, images, knowledge bases etc; requiring full and open access to spatial information.

A characteristic of these sophisticated spatial analysis tasks is that they require selective access to small elements within a large and seamless spatial database. Or they may require parallel processing across many machines on a network. One example is facilities management where it is desirable for the corporate data to be centralised, yet enable concurrent access by planners and estimators. Multiple versions of plant must be handled within different proposals and scenarios—often overlapping the same spatial location. All of these tasks that make full use of spatial information for the enterprise, require open, seamless access and interoperability.

Technology Drivers

We must look at the technology drivers within mainstream IT and examine how they can be applied to opening spatial information systems. The three main technology drivers enabling the mainstream IT revolution are: software, databases and communications.

Software

Modern software technology aims at maximum maintainability, reusability, applications migration and support of evolutionary development. Proprietary languages available within spatial information products have varying degrees of success in addressing the structured or object oriented methodologies typically used within modern software engineering. Rapid prototyping or evolutionary development has also met with mixed success within spatial information products, but is generally acceptable. By far the biggest drawback with proprietary languages is in the area of applications migration. Your application is wedded to the particular vendor package for good or bad. There are also problems with steep learning curves for new developers and the developers' reluctance to isolation within a non-industry-standard language. The next generation of spatial information systems packages should not attempt to reinvent the wheel by insisting on proprietary languages for applications development and customisation. There are plenty of good languages out there which can perform the task. An interesting example of this concept is the windows-based spatial viewer SDV. All objects within SDV, including documents, windows and the spatial objects themselves, are exposed to the outside world using OLE automation. Thus applications that manipulate SDV views and the objects within SDV may be written using any third party language which supports OLE. Some of the most widely used examples are Visual Basic, Visual C++ and Borland Delphi.

Databases

Database technology is the second pillar upon which enterprise data access rests. Modern relational databases (RDBMSs) are unequalled in their data management, data integrity and security features. Databases provide rapid and efficient data access, support of concurrency and distribution of data across the entire enterprise. Structured query language facilitates access to entities within the database, at any level of granularity and without concern for the storage details of the data. All these attributes are highly desirable for spatial information systems, except that relational databases do not support spatial data types, nor indexing of those spatial data types.

This is a general problem in some areas of the IT industry where the relational model falls short of providing complex or abstract data types. Two diametrically opposed solutions have been proposed to address this problem. The object database camp has opted for object based database management systems (OODBMSs) which store and retrieve persistent instances of user defined object classes. OODBMSs have the advantage that complex or abstract data types such as spatial data are natural parts of the architecture. A major drawback of OODBMSs is their comparative immaturity and miniscule installed base. The vast majority of legacy and corporate data in the world today, resides in RDBMSs. From the point of view of bringing spatial information systems into the enterprise, spatial data in an OODBMS is little better than using proprietary file storage.

The other strategy for handling spatial data within databases is to extend the relational paradigm to incorporate complex or abstract data types. The SQL3 standard is an attempt to bring abstract data types to the next generation of relational databases. In addition, some RDBMS vendors have started supporting spatial data types along with other abstract data types such as video and images. The major problem with these vendor supplied data types is that they simply are not mature enough to be useful for spatial information systems.

SPATIALinfo's Spatial Data Manager (SDM) is a spatial RDBMS that extends an industry standard RDBMS to include spatial data types and spatial qualifiers. Full spatial integrity and topology is supported. Utilisation of RDBMS technology enables a seamless mapbase with infinite granularity. Integration of spatial and non-spatial data in the enterprise is achieved by utilising and extending the RDBMS technology upon which mainstream IT is already based. Spatial information systems come into the enterprise without affecting existing data or applications.

Communications

Distribution of resources throughout the enterprise relies on fast and reliable communications infrastructure. Local and global communications technology is possibly the fastest growing area within mainstream IT. Fast communication is becoming cheaper and more readily available. Client-server applications and distributed databases are commonplace within mainstream IT. Most communications traffic, today involves relatively low volume, fixed record traffic for text-based applications. The advent of higher volume and less well structured traffic for multi-media and spatial applications will test and ultimately expand the capabilities of today's communications hardware and protocols.

The nexus of communications technology and object oriented technology lies in the distributed object architectures of which CORBA and Microsoft (OLE/COM) are high profile players. From the view of traditional IT, this is an upside-down world where data utilises an application, rather than the application owning the data. Data are free to roam the network, calling upon the services of applications as they go. Compound documents may encapsulate many layers of data instances. Electronic document management (EDMS) technology can be used to control the flow of documents around the enterprise to form a workflow solution.

SPATIALinfo and Spatial Information Systems

The spatial information systems industry is in the midst of a paradigm shift as we come out of the proprietary, monolithic dark ages and join the mainstream IT community. SPATIALinfo is committed to driving this paradigm shift with the adoption of new technologies to build the next generation spatial information systems.

Our cornerstone spatial database server system is SDM that enables the corporate database for development of open spatial information systems. At the front end, GDS (or third party open systems compliant display engines) provide an infinitely configurable view into the database with support for interchangeable application modules encompassing AM/FM, GIS and AES.

In the very near future, a spatial information application may look something like this. A single workflow document contains a network maintenance project comprising a work ticket, bill of requirements and a configurable view into a version managed spatial relational database. The document traverses the enterprise via a predefined workflow path, accumulating project requirements documents, one or more network design proposals, signoff on a design and issue of a materials request to the (plant) warehouse computer. After completion of work in the field, the document is updated for as-built modifications to the proposal, followed by final signoff and database commit. Meanwhile a customer connection request launches a query into the database for available plant at a particular address. This action raises a work request for more network, or a work request for customer connection to existing network.

This is just one example of how workflow technology and spatial information systems can work together within the enterprise. The key to the usefulness of such as system is that all information, spatial or otherwise, is located within the corporate database and readily accessible from all parts of the enterprise. Access to spatial and non-spatial information is freely available to business analysts and forecasters, designers, maintenance, customer services and marketing. All have access to the most complete and up-to-date information. Similar systems using GDS display technology and SDM database technology are currently being implemented for a number of large telco and utility enterprises. The technology is out there and working.