Mark Diehl

MarkDiehl.com

Health Information Architecture, Data Modeling, and Enterprise Architecture Planning

INFORMATION MODELING PAST, PRESENT AND FUTURE

As large mainframe systems proliferated in the 1960's, the role of the computer systems analyst began to be defined. Initially derived from the industrial systems analyst, the efficiency expert who designed and optimized manufacturing processes, the computer systems analyst also focused on designing and optimizing processes within the system. The computer systems analyst used a variety of graphical devices. Initially this was very much an art form that evolved into a generally accepted approach and was disseminated through trade and technical publication and professional conferences. Modeling per sé did not exist during that time - what we now consider as modeling is better described as diagramming. It had a distinctly process engineering flavor and used graphical notations, predominantly flow charting.

During this early period the automated process was the focus of design attention, with function-centered analysis and design. Databases as we now know them did not exist - programs interacted with stored data typically using direct access or sequential access technology. Data structures were fairly elementary, with data elements mapped by position within a fixed-length record. IBM's launch of its Information Management System (IMS) hierarchical database in the mid-1960's signaled a shift in systems theory. During the 1970's the advent of the network databases added a layer of complexity. In hierarchical and network database systems, the DBMS was initially employed as an improved Indexed Sequential Access Method for data storage and management. Hierarchical and network database designs often resembled sequential file designs. Data was still directly tied to a specific program, and not considered a shared resource.

As experience with hierarchical and network databases increased, information management theory advanced in the mid-1970's and data-centered analysis and design emerged. The advantages of sharing data among programs and eventually throughout the organization became apparent. Codd's watershed article advanced relational theory, and the Pick and Mumps systems provided niche alternatives to hierarchical and network DBMS, and the object-oriented paradigm proposed an alternative view of systems and information architecture.

Through the 1970's and 80's systems evolved and data complexity increased markedly. Complexity led to an explosive growth in system cost and increasing risk of failure - together increasing the need for improved analysis and design approaches. Diagramming techniques proliferated. The information processing flow chart became formalized along with data flow diagramming, and a number of approaches for diagramming hierarchical data structures. Information modeling emerged as a discipline in its own right as diagramming techniques and notation systems coalesced into standards and industry best practices during the late 1980's.

Another result of the increasing complexity of systems and the development experience was a growing emphasis on system life cycle management and supporting methodologies. Modeling became an integral part of most of these methodologies and used primarily for analysis, design and communication purposes. More importantly, the linking of data analysis and design to functional analysis and design became the crucial element in ensuring that the system actually meets business needs. Functional modeling (e.g. data flow diagramming through business process modeling) became loosely coupled with data modeling. Functionality of tool suites at that time (e.g. Logic Works BPWin and ERWin) began to pass functional model parameters automatically to the data modeling tool. CASE tools provided alternative notation systems, and included schema options to support both transactional and analytical system designs. This function-data linkage, along with improved functionality and appearance of these CASE tools marked advances through the 1990's.

The growth and evolution of information architecture and modeling practice, and especially the supporting CASE tools, can only be envisioned accurately for the short term future. Architecture and modeling theory will hopefully remain ahead of the curve, while its application to real systems and the capability its CASE tools will likely track closely with surging advances in systems operation, management and analysis and design. There is an ongoing interest in a "Grand Unification Theory" where some expect a merging of structured (relational) and O-O paradigms. (Despite the fundamental differences in the paradigms, some believe the relational construct is the ground state of the OO construct.) Also likely over the next few years will be a new crop of CASE tools with improved functionality (e.g. broader selection of notation system), tighter integration of functional and data modeling, and easier cross-tool exchange of model content (e.g. transfer of full model content across dissimilar tools). This future will be exciting for information architects and modelers.