Evolving Centralised Legacy Systems to Intranet-based Architectures

Evolving existing systems to support business process change has been given a new dimension by the development of Internet technologies. These technologies offer the opportunity for platform-independent distributed working and allow organisations to consider radically different ways of accessing and using their information resources. These technologies are 'open' and solutions are pieced together from heterogeneous components from different sources. Organisations are installing Intranets (internal information systems based on Internet technology) to access to corporate information using WWW browsers. It has been estimated that the rate of Intranet growth is more than 100% p.a.[1] and more and more organisations are considering how to apply Intranets cost-effectively to support their business processes.
To do so, legacy system issues must be taken into account and this is the principal theme of this project.

 1.1 Objectives

 The overall aim of this work is to investigate some of the issues involved in evolving legacy systems, based on centralised mainframes, to Intranet-based architectures. Our key objectives (Table 1) are to address those problems in the requirements, design and management of Intranet-based systems which stem from existing legacy systems.

 Objectives & Success Criteria

 Objective 
To develop an approach to business process analysis which helps re-engineer processes to make effective use of Intranet-based systems.
Success Criteria 
Trial use of method which we propose by the process reengineering team of our industrial partner. Publication of results in journals or international conferences. 

Objective 
To investigate design problems of changing centralised information systems into Intranet-based systems with distributed data and functionality.
Success Criteria 
Successful implementation and demonstration of test applications showing how system models of Intranets may be used to support end-user development and the evolution of a character-based UI to a WWW-based interface. Publication of results in journals or international conferences. 

Objective 
To investigate how current techniques for system configuration management must be adapted for distributed Intranet-based systems.
Success Criteria 
Publication of a configuration management (CM) model in journals or international conferences. Demonstration of software support based on existing CM and Intranet software. 

1.2 The problem

 Many organisations are being forced into radical change in response to new competitive pressures and customer demands. The financial sector has changed particularly quickly over the past few years. All major banks are in the throes of a cultural shift where a bank branch is changing from an administrative centre to a marketing organisation, selling the bank's range of products. Back-office processing is being taken out of the branches and into centralised service centres dedicated to single functions such as commercial loans, mortgages, etc.
To support these process changes, our partner in this project, NatWest Bank plc, has decided that the best way to provide shared access to corporate data is through an Intranet front-end with software running on a standard platform on all desktops. This will provide a framework for coordination and information sharing across service centres. Some of the critical problems which NatWest recognise in making this transition are:

 ¥ Technology impact assessment In re-engineering processes, it is difficult to assess the complex relationships between existing processes and legacy systems and to assess how the use of Intranet-based systems should influence the re-design of these processes.
¥ End-user development An individual service centre must develop its own software for its own products so that it can respond quickly to customer demands. There is a lack of tools and techniques to allow the flexible interrogation and use of information in databases and legacy systems to meet the demands of a shift towards a responsive organisation.

 1.3 Required research

 1. Requirements analysis Understanding how existing processes are constrained by legacy systems and how process re-design should take into account the use of Intranet technology. 
2. System design For end-user development, system modelling techniques are needed so that end-users can understand the distribution of data and functionality in the system and find the required functions and information to include in their application. Clear guidance is required on how to evolve character-based user interfaces from legacy systems to web-based interfaces using technologies such as Java. 
3. System management Understanding the configuration management requirements of end-user developed, Intranet-based systems. Centralised support is impractical and management strategies for distributed systems are required.

 1.4 Programme of work

 The work which we propose will be organised into three themes:

 Requirements

 What research will be carried out? 
There are two strands to the research which we propose:
¥ An understanding strand which will help us understand, in detail, the key problems which should be addressed by the technical research work of the project. ¥ A development strand where we will derive a systematic approach to business process analysis. This will be designed to help understand the impact of legacy systems on process change proposals and must be usable by organisation managers responsible for planning process evolution.

 Why do we need to study existing processes? 
We have already studied some NatWest business processes so have an understanding of their role in the organisation. We need further studies to assess the influence of legacy systems on these processes and the potential impact of a move to Intranet technology. The approach to process studies which we will develop from our work will help organisations avoid problems which arise when processes are redesigned without due regard for technology use. 

What will be the results of this research?
¥ A set of process descriptions plus associated commentary which helps us understand how legacy systems are currently used 
¥ A systematic approach to process analysis which assesses the use and impact of legacy systems and the influences of Intranet technologies on processes.

How will the work be done? 
¥ We will carry out an ethnographic study of business processes at NatWest to assess the impact of legacy systems on these processes. We will build on process studies that we have already carried out at NatWest but which have not considered the legacy systems issue. The study will aim to answer the following questions about the existing processes in the bank:

 Which aspects of legacy system use must be retained, as they are, in new processes with new systems?
Which aspects of legacy system use must be retained but which may change in new processes?
Which aspects of legacy system use are an accidental consequence of the process used and may be discarded in any redesigned process?
Which aspects of the process are a consequence of limitations of the legacy system and should be supported in a different way?
How do the inherent characteristics of Intranet-based systems constrain process re-design?

 ¥ We will propose a method for business process analysis based on our process studies and work concerned with integrating human factors analysis into requirements engineering. The above questions are, we believe, critical and the key objective of our technique will be to provide a means of finding answers to them. Currently, we favour the notion of critical process viewpoints (a legacy system may be a viewpoint). Process information is collected and analysed from a range of viewpoints. Some initial work in this area has already been reported [2, 3] but this was not concerned with the process impacts of legacy systems. 

¥ We will test our approach by using it for further studies at our collaborator's sites. Based on our assessment of its strengths and weaknesses, we will develop and improve it so that it can be integrated into NatWest's current techniques of process analysis.

 1.4.2 System design

 What research will be carried out?

 ¥ The modelling of Intranet-based systems We will develop guidelines for modelling Intranet-based systems and demonstrate how these models may be used to facilitate the development of tailored, end-user developed, front-ends to legacy databases. 

¥ The evolution of character-based interfaces to web-based interfaces We will develop a set of guidelines and procedures for converting the character-based interfaces of legacy systems to graphical user interfaces embedded in a WWW browser. These will help users decide how to convert character-based representations to web pages and provide a basis on whether to include functionality in the client, in the web server, in the legacy system or in helper applications used for data manipulation and presentation.

 Why focus on these design issues?

 End-user system development essentially involves collecting 'raw' data from a database then using a collection of tools (which may be embedded in web browsers as helper applications) to view, manipulate and present that data. It may involve some programming or reuse of programs developed for other applications. To develop systems, end-users need to know:
¥ What data is available and how it is organised. Remember the database may be hidden behind some legacy system so not all data as defined in the schema is necessarily available.
¥ What functionality is available for reuse either in the legacy system, in the Intranet server or in the client?

 To maintain these systems, we must also know what specific software has been developed (e.g. Excel macros) and how this is linked to other functionality in the system. Therefore, providing models for end-users to visualise the Intranet-based system using software available on their desktop is essential for effective end-user system development

 Most legacy systems have character-based interfaces. In an initial migration to Intranet-based systems, these can be emulated but, in the longer term, there will be overwhelming user demand to convert these to graphical interfaces embedded in a web-browser. However, interfaces often embed business rules in the form of validation checks and there may be complex sequences of inter-dependent screen presentations. This functionality must be moved from the legacy system to either the user's client or the Intranet server. We need ways of presenting the interface structures and rules and guidelines to help distribute user interface functionality. We also need to address the issue of how to make effective use of languages, such as Java, in migrating these interfaces.

 What will be the results of this research?

 ¥ A set of model types for visualising the data and functionality of distributed Intranet-based systems, notations for describing these models and demonstrators showing how these models may be created, viewed and used to facilitate end-user development.
¥ Techniques for modelling user interfaces and set of guidelines for user interface evolution which take into account the need to distribute interface functionality between web browser scripts, helper applications, the Intranet server and the legacy application. How will the work be done?

 System modelling
¥ We will identify a set of end-user applications which may be developed using data in a legacy database. As discussed in 2.6, it isn't practical to use real banking systems so these will be based on internal University legacy databases. We will develop several of these applications as demonstrator systems using our internal Intranet.
¥ We will analyse these applications to identify the structures used (e.g. the system architecture) and will develop a set of models of the distributed data and functionality.
¥ From the knowledge gained, we will propose a technique for modelling Intranet-based systems. We will refine this using other end-user applications which we have developed.
¥ We will investigate how the models of applications may be reused and applied in the development of new end-user applications.
It isn't cost-effective to buy CASE tools for modelling nor to train users in the use of these tools. We will ensure that all of the models which we propose can be created, manipulated and viewed using helper applications embedded in a web browser.

 User interface evolution
¥ We will investigate techniques of user interface modelling which show the different screens used in the system, their structure, the dependencies between these screens, transitions from one screen to another and associated field validation rules. We will develop these using our trial applications.
¥ We will carry out a number of small experiments, using local systems (see Section 2.6) to allow us to build up an understanding of the best way to distribute interface functionality between the existing application, the server, scripts in the client or helper applications. 
¥ Based on this experience, we will develop a set of guidelines for modelling character-based user interfaces and adapting these for including in WWW browsers.

 1.4.3 System management

 What research will be carried out?

 ¥ We will consider one key problem area namely configuration management. We will investigate the implications of distributing functionality between legacy systems, Intranet servers, client tools and WWW browsers for configuration management policies and will propose new Intranet-based approaches to distributed configuration management. 

Why is system management an issue?
System management is probably the single largest cost in managing PC networks and Intranet-based systems present new management challenges. There is a clash between the need for global system management and the desire of end-users to tailor the functionality of their system to meet their specific needs. Currently, legacy systems are managed by IT professionals who are aware of the problems of hardware and software installation, configuration management, data security and integrity, etc. The evolution to Intranet-based solutions will require much more end-user involvement in this process and an evolution of the current centralised models and configuration management processes.

 What will be the results of this research?
¥ A configuration management model which may be used to design configuration management procedures and associated support for the evolution of Intranet-based systems. We will produce a simple demonstration system to demonstrate the feasibility of the approach used in our model.

 How will the work be done?
The work on developing a configuration management model will be addressed in three interleaved phases:
¥ A study phase where existing management processes will be studied and documented.
¥ An analysis phase where we will look at these existing processes to discover what must be retained in any distributed configuration management solution and where we will assess the requirements for distributed configuration management.

 ¥ The development of a configuration management model which will allow configuration entities to be identified and located and which will propose configuration policies which may be adopted for applications where the functionality is distributed between server code, client code and user-generated scripts.

 1.5 Scientific/technological relevance

 The principal technical contributions of the work which we propose will be in two areas:
1. Business process analysis Our focus on the impact of legacy systems on processes is not addressed by other approaches which assume closed and centralised system support.
2. Software evolution There is little or no published research which has addressed the issues of Intranet-based system modelling, user interface migration or configuration management strategies for web-based systems.

 Suchman [4] and others, including ourselves [5-8] have found empirical studies to be valuable in understanding the real requirements of processes for systems support. However, none of these studies have considered legacy systems and system evolution to support process reengineering. There has been some work on developing methods for process studies such as Contextualised Enquiry [9] and our own work on 'quick and dirty' ethnography [10]. However, for an approach to process analysis to be viable, it must be usable by managers in the business itself and no other approach addresses this issue.

 The term Intranet was supposedly invented in 1995 so there have been few, if any, research results which have specifically addressed the issues of transition to Intranet-based systems. Gregory and Reingruber [11] have briefly discussed reengineering with Intranets but, we suspect, without practical experience in this area. Various commercial products such as that available from Amazon [12] support the development of Intranet front-ends to legacy databases. These products do not really address the problem of converting the functionality of existing legacy systems into Intranet-based solutions but simply provide a means to write new interfaces to existing company databases. Pirolli et al [13] discuss some of the problems of WWW visualisation and propose a technique for visualising sets of WWW pages. While this is not directly applicable to our work, some of their proposals may be adaptable to more general system modelling. Merlo et al [14] have investigated the problems of reengineering character-based user interfaces to GUIs. They developed a semi-automated approach based on attribute grammars and highlighted the importance of modelling the legacy system interface. Rice et al [15] have done work on developing web-based interfaces which is relevant but their focus was on new system development rather than the transition from character-based to graphical interfaces.

 1.6 Our research process

 Studies of the industrial partner's environment will be used to identify critical problems and key requirements for transition to Intranet-based systems. The study results will focus other research activities. The analysis method will be validated by use on further on-site studies of configuration management which will themselves provide important data for our research. 

Because of the confidentiality of the data and systems used by our partner, it will not be possible for use to experiment with these systems in our work. Rather, we will install our own Intranet server and will use University administrative systems as the testbed for our research. These systems for student and departmental management are typical legacy systems (they have character-based interfaces, are poorly documented, run on a large central system, etc.). Whil the processes of using these systems are clearly distinct, we believe that the technical problems of evolution to Intranet-based architectures are likely to be common to all systems.

 This approach has the advantage that it will ensure the generalisability of our research results beyond the immediate demands of NatWest Bank. We are confident that our results will be transferable to any type of legacy system migrating to an Intranet-based architecture. 

. References

 [1] Campbell, I., The Intranet - Slashing the Cost of Business1996, International Data Corporation: http://home.netscape.com/comprod/announce/roi.html/roi1intranj.
[2] Sommerville, I., Sawyer, P., Kotonya, G., and Viller, S. "Process Viewpoints". in Proc. 5th European Workshop on Software Process Technology. 1995. Amsterdam, The Netherlands. 
[3] Hughes, J., O'Brien, J., Rodden, T., Rouncefield, M., and Sommerville, I. "Presenting Ethnography in the Requirements Process". in Proc. RE'95. 1995. York. 
[4] Suchman, L., Plans and Situated Actions. 1987, Cambridge: Cambridge University Press. 
[5] Heath, C. and Luff, P. "Collaborative Activity and Technological Design: Task coordination in the London Underground control room". in ECSCW'91. 1991. Amsterdam. 
[6] Bentley, R., Rodden, T., Sawyer, P., Sommerville, I., Hughes, J., Randall, D., and Shapiro, D. "Ethnographically-informed Systems Design for Air Traffic Control". in CSCW'92. 1992. Toronto, Canada. 
[7] Goguen, J. and Linde, C. "Techniques for Requirements Elicitation". in RE'93. 1993. San Diego, Ca. 
[8] Anderson, R.J., Hughes, J.A., and Sharrock, W.W., Working for Profit: The Social Organisation of Calculability in an Entrepreneurial Firm. 1989, Aldershot: Avebury. 
[9] Holtzblatt, K. and Beyer, H., "Making Customer-Centred Design Work for Teams". Comm. ACM, 1993. 36(10): 93-103. 
[10] Hughes, J., Rodden, T., King, V., and Andersen, H. "Moving out from the control room: ethnography in system design". in CSCW'94. 1994. Greensborough, North Carolina. 
[11] Gregory, B. and Reingruber, M., Reengineering with Intranets1996, Enterprise Reengineering: http://reengineering.com/articles/may96/mkgwork.htm. 
[12] Shafe, L., Amazon - Web to Legacy1996, Intelligent Environments: URL http://www.ieinc.com/cssumm.htm. 
[13] Pirolli, P., Pitkow, J., and Rao, R. "Silk from a Sow's Ear: Extracting Usable Structures from the Web". in Proc. CHI'96. 1996. Vancouver, Canada. 
[14] Merlo, E., Gagn*, P.-Y., Girard, J.-F., Kontagiannis, K., Panangaden, P., and De More, R., "Reengineering User Interfaces". IEEE Software, 1995. 11(1): 64-73. 
[15] Rice, J., Farquar, A., Piernot, P., and Gruber, T. "Using the Web instead of a Window System". in Proc. CHI'96. 1996. Vancouver, Canada. 

Some 'legacy' links:

 There are several sites giving basic definitions of legacy systems and approaches to the legacy problem.

 The Renaissance Project contains a whole host of links to topics related to 'legacy' issues including articles on 're-engineering' legacy problems. 

Amonst the papers it highlights is one entitled Legacy Systems Triage Renovation, Re-engineering, or Redundancy?. which suggests that the fundamental problem facing Legacy Systems Management is that of triage, determining which shall live, which shall die and which shall benefit from the available resources. 

Some Intranet links:

 There are numerous sites that discuss the effectiveness and business case for the use of Intranets' the Intranet Resources site points to a number of these while other articles on Intranets can be found here

 One interesting paper on Intranets is "Internet Technologies Deployed Behind the Firewall for Corporate Productivity" discusses the opportunities for deploying Intranets - Internet technologies used on the corporate network - for a variety of strategic and tactical communications purposes and provides a series of recommendations on the business, social and organizational issues involved in rolling out an Intranet site.