Time to Mobilize
- Application Migration
- Database Migration
This report explains how automated software transformation works and proposes a strategy for assessing the suitability of existing applications for migration to modern platforms such as J2EE and .NET.
Since 1988 the Technology Research Club (Club de Investigación Tecnológica) has been carrying out applied research on the applications and implications of IT in organizations in Costa Rica. To date the Club has sixty member organizations, the most forward looking and successful private and public organizations in the country. This is the first report written in English and to be made publicly available, it is the 33rd research report. For more information on the Club, its members, the latest activities, as well as a complete list of research reports, visit www.cit.co.cr.
A legacy application is any application based on older technologies and hardware, such as mainframes, that continues to provide core services to an organization. Legacy applications are frequently large, monolithic, and difficult to modify, and scrapping or replacing them often means reengineering an organization's business processes as well. Legacy transformation is about retaining and extending the value of the legacy investment through migration to new platforms.
Re-implementing applications on new platforms in this way can reduce operational costs, and the additional capabilities of new technologies can provide access to valuable functions such as Web Services and Integrated Development Environments. Once transformation is complete the applications can be aligned more closely to current and future business needs through the addition of new functionality to the transformed application.
In short, the legacy transformation process can be a cost-effective and accurate way to preserve legacy investments and thereby avoid the costs and business impact of migration to entirely new software. This report explains how transformation works and proposes a strategy for assessing the suitability of existing applications for migration to modern platforms such as J2EE and .NET.
The goal of legacy transformation is to retain the value of the legacy asset on the new platform. In practice this transformation can take several forms. For example, it might involve translation of the source code, or some level of re-use of existing code plus a Web-to-host capability to provide the customer access required by the business. If a rewrite is necessary, then the existing business rules can be extracted to form part of the statement of requirements for a rewrite.
The report takes the view that J2EE or .NET are suitable target platforms for transformation. The arguments in favor are based on technical and cost factors, on the fact that most automatic translation products target these platforms, on a growing skill-base in J2EE and .NET, making it easier to recruit staff, and on the availability of standard XML-based protocols for use by other applications, which facilitate the publication of application function to a network (usually referred to as "Web Services").
Substantial automation of this transformation process is now feasible, making transformation an economically attractive proposition compared with rewriting or replacing the legacy application. The available tools cover all aspects of the process, although some manual intervention will be required. Using the tools in practice will depend on the scale of the task and whether automation is necessary or economic in every case. It is assumed of course that the existing applications are of sufficient quality and fit business needs well enough to make them worth transforming.
The tools available in the market are point solutions — they are applicable to specific scenarios and only handle part of the transformation process. Consequently there will be a need to buy in services to help design and execute the transformation as there are too many unknowns to be overcome without help from experts with previous experience. In-house resources will be needed to impart available knowledge about the legacy applications, and to build up the future knowledge base for maintaining the transformed applications and aligning them with business needs.
Transforming legacy applications is a task with both risks and rewards. It is easy to fall into the trap of relying on what seem like stable applications and hoping that they will be adequate to keep the business going, at least in the medium term. But these legacy applications are at the heart of today's operations and if they get too far out of step with business needs the impact will be substantial, and possibly catastrophic. The challenge for the CIO is to present the arguments for the legacy investment in the best possible light, but also to give management the full picture of these risks and rewards so that they can make a decision in full possession of the facts. Ultimately, legacy transformation is an "enabling" project, that allows other things to happen, but has its own direct benefits as well.
In selecting a supplier or suppliers for a transformation project, it is best to strike a balance between the project-oriented players (who will take care of the transformation itself), and the infrastructure suppliers and in-house staff who have to make the end-result work every day. In today's state of the art, transformation expertise must be at the heart of the solution delivery. This can be done by appointing an independent project manager (in-house or contractor), and keeping functional changes and integration work separate from the tasks of code translation, data migration and associated testing.
In conclusion, the automated tools and techniques now available make legacy transformation technically and economically feasible. Replacement or rewrite are necessary in certain instances, but if the existing legacy application meets current business needs and the quality is good, then the chances are that the legacy asset can be effectively transformed to continue to meet the needs of the business in the future.
A legacy application may be defined as any application based on older technologies and hardware, such as mainframes, that continues to provide core services to an organization. Legacy applications are frequently large, monolithic and difficult to modify, and scrapping or replacing a legacy application often means reengineering an organization's business processes as well.
This report aims to explain the legacy transformation alternative, which maintains and extends the value of the legacy investment through migration to new platforms, and at the same time limits the need to reengineer existing business processes. It includes proposals for a legacy strategy and discusses transformation planning and cost justification issues. The report's intended audience include CIOs and their direct reports, systems integrators, and suppliers of commercial off-the-shelf application packages based on older technologies.
Chapter 2 considers the basic motivation for legacy transformation, and a description of the process follows in Chapter 3. Chapter 4 provides guidance on how to make a decision about when to go for transformation, and when to scrap or replace the legacy application (or in some cases, when to invest further). In particular, it reviews the difficult question of choosing a target platform. Chapters 5 and 6 deal with project planning and business case development, respectively. Chapter 7 discusses the supply-side options and makes some recommendations on choosing business partners to assist with transformation. Finally, Chapter 8 provides a general outlook for transformation. The Appendix contains a brief glossary of abbreviations and terms used in the report.
Existing applications are the outcome of past capital investments. The value of the application investment tends to decline over time as the business and the technology context changes. Early in the life-cycle there will be enhancement investments to maintain close alignment with the business but eventually there will come a point where this becomes difficult. This can happen, for example, where the underpinning infrastructure is superseded, web access is required, or the weight of changes in the applications and lack of available know-how make it impossible to continue with enhancements.
Dissatisfaction with legacy centers on inflexibility (takes forever to make changes, can't make changes), maintainability (no documentation, no-one understands it, lack of skilled people), accessibility (can't make it available to customers, for example), cost of operation (runs on costly mainframe infrastructure, high license fees), and interdependency of application and infrastructure (can't update one without the other).
At this point we have a choice: do we initiate a process of renovation and transformation, or do we write the application off and find a replacement?
Re-implementing applications on new platforms can have benefits through reduced operational costs, and through the additional capabilities of new technologies it provides access to valuable functions through more economical means. Migration to a new platform also provides an opportunity to align applications with current and future business needs through the addition of business functionality and through application restructuring.
Drivers for legacy transformation are operating cost reductions, mergers and acquisitions, internal reorganization, new corporate infrastructure, need for Web-enablement, outdated performance and functionality, data consolidation, and positioning for future changes such as B2B via XML and SOAP (Web Services).
Web Services may become a key factor in forcing change on legacy applications. Most organizations are in what might be referred to as the "phase 1" stage of Web Services planning — perhaps running trials or simply assessing how important this concept will become in the future. Some are at "phase 2," and are already exploiting the integration capabilities, often internally in their organization. Realistically, Web Services could become a strategic issue in the near term, adding urgency to the need to take action on the legacy applications that will be at the heart of the Web Service infrastructure.
It is the position of this report that the tools and techniques to support automated transformation are now such that migration is both technically and economically feasible. Replacement and rewrite are necessary in certain instances, but if the existing legacy application meets current business needs, then the chances are that this legacy asset can be effectively transformed to continue to meet the needs of the business in the future.
This section looks at current approaches to transformation and the tools available to help make it happen. It provides a model framework to assist in distinguishing between the wide ranges of products available in the market.
The basic requirement for a successful legacy transformation is to retain (and add to) the value of the legacy asset. In practice this transformation can take several forms, for example it might involve translation of the source code, or some level of re-use of existing code, with the addition of a Web-to-host capability to provide the customer access required by the business. It will be assumed throughout the discussion that the goal is to move to a commodity/open platform (such as J2EE or .NET) and that some additional functionality may be added in the process.
The diagram below is a model of the legacy transformation process including all the key activities involved. The sequence of activities can vary and there will be an iteration through the process when transforming a portfolio of applications, for example. It is usual practice to complete existing code translation, data migration and associated testing before adding new functionality. This is to test the end-result for equivalence with the existing application and to prove correctness of the translation and data migration before any changes are made to program logic and structure.
The input to the process is a legacy application and the outcome is a transformed application with most of the legacy asset intact, possibly enhanced, and integrated into the overall applications portfolio of the business. The process is subdivided into the several process steps necessary to bring about this transformation. It should be understood that, like most process models, there may be several "paths" through these process steps, and there may be some iterations. For example, it may be decided to re-use some legacy code by "wrapping" it in Java, while some other code is translated directly into Java — this involves different routes through the sub-processes. An example of iteration is when the transformation happens in stages: One part of the application is transformed to the point where it is integrated and goes live; then the process starts again to transform the next part of the application; more integration is required; and this process is repeated until all parts of the application have been transformed.
This logic also applies to suites of applications that share the same data or deliver a common business function. Here it is best to transform the suite as a group, but if this is not feasible, then additional programming will be needed to provide the "scaffolding" (adapters) required to keep the suite of applications operational during the transition phases. Typical activities making up these process elements are listed in the table below.
|Analyze and assess the legacy applications||
|Set transformation goals and success measures||
|Administer and control the transformation||
There is now a range of tools, products and standards available to assist with the transformation process as described above:
These tools, products and standards:
In considering the feasibility of automated tools for transformation, it is useful to consider this list of obstacles:
Specifically, this is how the obstacles listed above are addressed:
|Monolithic applications||Modelling the architecture of legacy applications separates out the client-tier, application logic and persistent data. Once migration of the code is complete, componentization can begin, to align the components with business function. Representation of applications and data as components enables rapid and easy assembly of new business functions. Current state-of-the art limits componentization to business functions as a general rule, and may be limited by the structure of the original code.|
|Application tied into OS-specific facilities||Translators include automatic translation of OS-specific program calls into the equivalent on the target platform. Where there is an unrecognized call, this is labelled as an exception, to be manually translated. If there are many occurrences a solution can be retrofitted into the tool by the supplier. An approach found in some translators is an "analysis" run on sample code before translation commences, to identify the likely number of exceptions. If these fall into a small number of categories, the translator can be adjusted to deal with them automatically.|
|Incompatible stove-piped applications||The great advantage of the modern open application platform is its ability to integrate diverse applications, once the initial conversion of the code has taken place to the target platform. The evolutionary development approach adopted by most businesses means that transformation will happen in conjunction with existing applications, data and infrastructure. Transformed applications will have to coexist and interoperate with those applications.|
|Code holds the business rules and no-one knows them||Application mining can separate interface code, flow control, IO, and so on, from the small proportion of the code that represents business logic, thus isolating the business logic. The search for business logic can be narrowed down by automated searches for code constructs that typically embody business rules. This is followed by code-inspection workshops with the code maintainers. A complementary approach starts with the data whose value reflects the execution of a rule and trace the code that sets the value (this is most useful in Y2K, Euro type situations but can be generalized.) From these the implicit processes are mapped and ready for input to the definition of requirements or new functionality. In parallel with this effort, a know-building process is undertaken (more about this later in the report).|
|Where are the real boundaries of the application?||A combination of application mining and manual inspection enables the boundaries to be mapped. Specific steps are taken in 4GL translation, for example, to deal with run-time libraries and similar infrastructure. Transformation methodologies specify that to ensure functional equivalence, all code, interfaces, job control and data need to be considered. This preserves functional integrity and reduces testing.|
|Code issues — duplication, un-executed code, identifying the "right" version of an interface, and so on||Sorting this out is a by-product of the application mining referred to above — recognizing however that there are situations where the requirement is to simply move the code from an older platform, without any attempt to establish the business logic. Automated tools for persistent data conversion can produce data element definitions, determine which source records are used for data in tables in the target database and designate the source for the data in each table.|
|Code quality issues||Translation tools can applied to a sample of code to assess it. If the issue is fundamental than that, for example, the application does not function correctly, then it may be necessary to drop it as a candidate for transformation and look to extract the business rules for a rewrite. Automated inspection tools exist (such as Advantage) to assess structure).|
The diagram below shows the process model with examples of current products mapped onto it. This is a representative selection only and there are other products available and other suppliers operating in this space. The list does demonstrate that available tools cover the whole process in one way or another. It should be noted that most of these products are "point solutions." That is, they typically deal with one or two specific programming languages, and they are restricted to a small number of target environments. For example, the translators are constructed to handle specific languages, although they can be extended fairly readily to handle other languages as required by the market.
Further, the products are generally limited to one part of the transformation process. This is an inevitable consequence of the structure of the transformation process, because although the transformation process is presented as a monolithic transformation, in fact the elements are quite different from one another, and the state changes are not at all consistent. For example, the Translate activities change code from one language to another, while the Analyze and Assess the Legacy Application activities change a state of (relative) "ignorance" to one of "knowledge." Thus it is reasonable to suppose that the tools that support automation of these separate process elements will remain distinct (even when they are accessed through a common interface, for example).
This is not necessarily a bad thing, but it does mean that more than one tool is likely to be needed and some mixing and matching will be necessary - with sufficient expertise and good project management, a transformation project will succeed. However, it does suggest that some skepticism is required when a supplier talks about a solution that "will take care of everything."
There are limits on automatic decomposition of applications into components that could become building blocks for new applications. In general, current transformation techniques take the applications to a new platform, not to a new architecture. It is not easy to manually re-structure the transformed application to create components below the business function level. A feasible and realistic goal might be a target architecture consisting of business objects that encapsulate the business logic of a single entity and the data particular to that entity. Ultimately, the feasibility of re-use depends on the structure of the original legacy application — well-shaped structured code lends itself more readily to some componentization, whereas linear, monolithic code needs wrapping in total.
A strategy for legacy transformation is discussed in the next chapter.
Developments in legacy transformation tools and techniques provide an opportunity for the business to review its legacy portfolio. To transform or replace, to scrap of re-invest? These are the questions to be addressed by a legacy strategy.
When the requirement is to transform a single legacy application, then the choice of which way to go will be decided by the quality of the application, and whether or not there are off-the-shelf products readily available to replace it, if the quality is poor. In most other situations the CIO needs to begin with more fundamental business questions.
There are four common drivers for considering a transformation project:
In practice some of these drivers may be combined. These drivers can be seen as a kind of spectrum, going from "push" factors (which mean that the legacy application must react to fit a new environment) to "pull" factors (the business wants to seize an opportunity to grow/reach new customers/add products and services).
These drivers are contrasted in the diagram, suggesting that there can be quite a variation in the value that the business will be looking for from a transformation. The following points need to be considered:
In any business there is the obvious distinction between financial, sales order processing, human resources and manufacturing applications, for example, but of course within these categories there are often several hundred individual applications with millions of lines of code between them. A pre-assessment of the portfolio is desirable before proceeding to make individual decisions about transformation methods.
The emphasis of the pre-assessment is on business value (one of the aims of transformation is to address technical quality, so this question is postponed until later in this process). The goal is to streamline the applications portfolio by reviewing the existing applications to see whether they continue to provide business value. If applications can be identified for decommissioning this will result in immediate savings. This streamlining will require commitment from the business. Here is a starter list of questions to ask about each application:
There are alternatives to total transformation of the legacy application, and at this point in the development of the transformation strategy these alternatives should be reviewed. The main decision factors are (I) the quality of the legacy application, and (ii) the availability of replacement packages. "Quality" in this case is a subjective term applied to the application itself, regardless of the platform it runs on or the code in use, and should be assessed in terms of such parameters as:
In summary, the "quality" assessment is about the suitability of the legacy application in business and technical terms.
The availability of a replacement package is the other important consideration and this will depend on the uniqueness of the current application. If the quality of the legacy application is poor and there is comparable functionality available in a third-party software package, it makes sense to replace it.
There are four broad options: transform, re-use, rewrite or replace. As explained earlier, some or all the elements of the transformation process apply to the first three. The diagram1 overleaf shows the combinations of factors that might lead to these four options.
Transform: Apply the transformation process, adding functionality and business reach as required.
Re-use: There are two possibilities here, one where the legacy application is centered on a third-party package/DBMS already, and the other where the business has developed its own application from scratch. If the legacy application portfolio is largely centered on a third-party package then the best way forward may be to upgrade to the latest version and use wrapping techniques to provide the required reach and other functionality improvements. For in-house applications consider wrapping the application. To provide direct access to data by end-users without going through the legacy application will usually mean adding a back-end data warehouse as well. The drawback of this approach is that it adds more elements to be maintained and two sets of data to be kept synchronized.
Rewrite: The key asset here is the business rules and data structures — the application is the problem. Application mining and analysis of code logic and data structures is required to provide the starting point for the rewrite.
Replace: Look for a suitable package or outsource. Be prepared to make changes to the business model to meet the package half-way.
The diagram below summarizes the overall decision process in graphic form.
Transformation projects are broadly aimed at converting code or modules to Java, C++ (or C#), to a relational database environment, or to a HTML architecture — or indeed some combination of these. The majority of new enterprise-level development in the foreseeable future will take place on one of two platforms: Microsoft's .NET platform or the multi-vendor Java 2 Platform Enterprise Edition (J2EE). This is not the same thing as saying that all legacy transformation should target one of these two platforms, but there are sound arguments in favor of doing just that:
The very factors that make these platforms suitable for Web Services are of course of great interest in transforming legacy applications, because of the way integration can be facilitated. Most legacy applications form part of an existing portfolio of interrelated and interdependent applications and Web Services are a next step in the evolution of application integration. Drawbacks include the continuing evolution of standards for Web Services and resolution of security issues.
The choice of .NET versus J2EE is the subject of much debate. Both have evolved from existing application server technology. J2EE is quite mature and is already running large-scale enterprise applications. .NET is the newcomer, but definitely here to stay. Today, Microsoft-based solutions are limited nearly entirely to Wintel class platforms — in other words, the choice of .NET implicitly chooses the platform, middleware and operating system and it is arguable that it takes more effort to scale to several hundred concurrent users that a similar J2EE implementation. Conversely, the smaller organization may choose one platform as their near-universal standard, and will go with .NET for its low cost of entry and focus on rapid application development. Both J2EE and .NET are being repositioned to deliver the Web Services vision.
The strategic arguments for each side are familiar: Platform portability versus vendor lock-in, cost advantages of a bundled supplier versus dealing with several suppliers, a better architecture versus the risk of instability while the architecture is implemented.
The bottom line is that either platform is a suitable target for legacy transformation and the choice between them is best made on the basis of business and technical strategies overall.
There may be considerations that lead to other target options. Some possibilities include:
A transformation project exhibits many of the characteristics of traditional development projects such as objectives setting, user involvement, testing, scheduling, monitoring, and so on. There are however some factors that differentiate a transformation project:
The classic development project follows a four-stage pattern of Plan, Analyze, Design, and Implement. The greatest risk is that the transformation project will be seen as implementation only, with little need for plan, analyzing, or design. One way of understanding the totality of the work involved is illustrated in the diagram below.
This is a logical expression of the tasks and activities involved, not a definitive way to complete the project. Any organization undertaking a transformation project should check its plans against this diagram to ensure that nothing has been forgotten. The diagram illustrates the point that initial work is needed before progressing the transformation project. For example, the available transformation technologies may have to be sourced from multiple suppliers, and it is likely that consulting expertise will be needed, at least for a first transformation project. Secondly, when there is uncertainty about the quality of the legacy application it can be worthwhile to run a proof of concept mini-project to confirm feasibility and (perhaps equally important) to firm up the total project costs. Subsequently the requirements for additional functionality can be acquired, following standard development procedures, with the proviso that this functionality is being added to an existing application, so the business users are not starting with a blank sheet.
Building know-how is emphasized here because of the peculiarities of the legacy asset. Although not always the case, legacy applications are typically lacking in documentation, business users are unfamiliar with the precise nature of the business rules built into their systems, and the technical expertise is confined to perhaps one or two people in the IS department. The Build Know-How activities might include the following:
The "method" described in the previous paragraph described a framework for transformation. The specific features listed earlier are partly addressed by this framework, and by additional specific activities:
It is advisable to involve a consulting partner, at least for the first project — someone who will anticipate the many "gotchas" and who knows their way around the target environment. Here's a checklist: Do they have a methodology? Have they access to the necessary tools? What is their track record? Do they know their way around .NET, J2EE or whatever target platform you plan to adopt?
Making adjustments to existing development methodologies — Transformation projects require structure and deliverables like any other project. Any methodology based on the V model will work effectively, although some modification to procedures may be necessary to follow the principles of user, architectural, and implementation model testing.
Waterfall methodologies fit the V pattern as do most "lightweight" methodologies.
Lightweight methodologies have superseded the traditional waterfall approaches in many development shops. These lightweight (agile) methods are adaptive and cope well with change, and are people-oriented. They are suitable for legacy transformation because of the way transformation projects are approached (it is necessary to get through the conversion phase of at least part of the existing portfolio before it can be confidently predicted what the introduction of new functionality will cost), the low effort required in design (the requirements are largely inherent in the legacy), and the proportionately higher effort required for testing and cutover. If a development department already has a documented lightweight methodology in place, some adjustments may be needed to fit a transformation project. For example, the scope and definition of the earlier phases and deliverables will need to be adjusted, and activities such as testing will need to be changed.
There is a common perception in business that legacy systems are yesterday's solutions. The legacy application is not seen as a basis for going forward, but as outdated, difficult to maintain, and possibly lacking a good fit with business needs. A champion is essential to promote the transformation option and put forward the business case.
This will not be easy, as there will be interests opposed to transformation and favoring other options. In-house technical staff have little incentive to keep the legacy application but will lean to a rewrite or package replacement options because these provide new development experience and the opportunity to learn new skills. Existing platform suppliers will want to keep the status quo. Senior managers will be at the receiving end of marketing campaigns from ERP and package suppliers, all promising benefits through replacement and process redesign. End-users may be more positive because the legacy application does the job they expect of it today, even though the cost may higher than they would like and it lacks certain capabilities (for example, no Web access). However, they will find it difficult to assess the technical advantages and risks of options that are quite different from one another (rewrite versus replacement by a package versus transformation).
Because of these considerations, it must fall to the CIO to develop the case for transformation and to explain the pros and cons of the various options to business managers. The CIO will have to push in-house analysts to look at all the options. Experience shows that they will prefer the design of a new solution over re-use of the old and this tendency needs to be counter-balanced by the CIO. Alternatively, a third party may have to be involved to see that an objective analysis takes place.
Note however that business managers need to champion the transformation project overall. It may appear to be a like-for-like replacement of the existing application, but there will be a need to accept the end-result as equivalent, and to be involved in any functional changes. And when there are problems or delays (as there always are), there needs to be someone to remind everyone of why the work is being done and the benefits that will accrue to the business as a result.
These are hard times when it comes to any form of business investment. The situation is made worse for legacy investments because many of top management's background assumptions and time-honored business models are inadequate to understand what is going on. The key is to find the right way to present the issue to top management, spelling out the downside of not investing, and contrasting the risks and benefits of the various alternatives.
What makes justifying a legacy transformation project so difficult? First of all, managers are most comfortable with the idea of spending money to get something new and at first sight transforming a legacy application seems like a project to "fix" something that already works. Second, the legacy project is often an "enabling" investment — that is, it positions the business to achieve something else. For example, one of the benefits might be to make it easier to add new functionality, implying that the functionality could be added some other, possibly less costly way. Indirect benefits of this kind are always more difficult to quantify and justify. Third, the costs of the transformation project have to be spelled out, while the true cost of today's legacy (disruptions, maintenance issues, costly operations, and so on) is hidden in other budgets. These are the issues covered in this section.
This is likely to be first reaction of a manager to a request to spend money on a legacy application. Like other situations in life, people become used to their legacy applications, the awkward interfaces, the workarounds, the time required to make changes, the cost of maintenance and operation, and so on. Every user is overworked and understaffed and people don't have the time to think about potential fixes and improvements. They may grumble and complain but they're too busy to do anything about it. So unless there's a crisis, they would just as soon get on with the job, thank you very much.
Management hears the grumbles and complaints but unless there's a demand from customers or moves by a competitor they are unlikely to react. Management is used to the size of the maintenance budget and it seems just as easy to approve the same budget for next year, without getting into an argument about it. Any proposal to transform a legacy application is therefore likely to get a negative response.
Overcoming this inertia is the first step in moving the legacy transformation proposal forward and to do this it is necessary to explain the risks and opportunity costs of doing nothing. For example, the risk posed by key maintenance staff leaving: Legacy applications by their nature are opaque, not-so-easy to work with, because of the layers of changes made over the years, and lack of clarity in documentation. (Documentation isn't usually created to help maintainers, it's there to help users, and to explain how and why the application was built that way in the first place.) So it takes long and close acquaintance to get to know the ins and outs of these applications, and hence the dependence on the people who maintain and enhance them. Of course people can be replaced. But this takes time, there's a learning curve involved, and if the programming language is relatively obscure, it may prove difficult to hire in the skills needed. Other potential risks include withdrawal of support by platform component suppliers and lack of replacement and upgrades for hardware and software.
At the business level there may be risks posed to future business plans still on the drawing board. Can the legacy application cope with those 50% extra customers? Will the people in Pre-Sales be able to cope with that many quotations, given the state of the application?
An example of opportunity costs is the staff time that could be released by moving to up-to-date tools. The current maintenance staff are tied up maintaining (with some difficulty) an ageing application - this work could take less time and effort on a transformed application due to the availability of tools and their inherent productivity. Changes and enhancements would be executed more quickly and efficiently, and the staff in question may be able to free up time for new development.
These arguments based on risk and opportunity costs (the costs of doing nothing) must be complemented by an explanation of specific benefits and how the legacy investment advances the organization's business goals. While these will be context-dependent, typical benefits can include long-term cost reductions, time-to-market improvements, and extended business reach, just to pick three.
After the investment push on legacy applications leading up to Y2K, management may have concerns about further spend on the same applications. The same reaction is likely to a transformation request for a two- or three-year-old "legacy" application. (This situation can arise, for example, when the build of the new application started some years back, but because of delays and time taken to roll out across the company, is already showing its age.)
In some ways this is a similar issue to the one in the previous section, but with an added dimension, as it implies there's a history to this legacy question. The basic issue here is that management has not been given a view of the future. In other words, their expectations have not been properly managed. This has its origins in the way we have treated large projects in the past, as relatively isolated, one-off events, rather than part of an on-going business program. Businesses have acquired the habit of thinking of an applications "project" as something that is finished when the application is delivered, the build team disperses, the consultants leave, and the users get on with the job.
The justification argument needs to focus on the treatment of software investment and get across the notion that software is (or should be) a non-perishable asset. Businesses have built up substantial investments in legacy applications and their effective operation is core to running today's businesses. It makes sense to keep a legacy application running if it continues to serve business needs. Yet the forces of business change, technology obsolescence and decay in internal know-how put applications at risk over time.
This is why business invests in maintenance and improvement. But is the legacy investment like a car, to be replaced by the latest model every so often, or is it more like a house, requiring regular attention and renovation, and an extension added if and when the family grows? Recent developments in tools and platform technology make it practical and sensible to take the latter view. Looked at in this light, legacy transformation becomes a way of prolonging the software investment and quite possibly a way to deliver better data and added flexibility for future expansion upgrades — as well as costing less.
Not every legacy justification starts from the same point. It is necessary to understand where this starting point lies in what might be described as the "Bermuda Triangle" of Cost, Value and Affordability. The common confusion and difficulty that IS finds itself in justifying investments in "enabling" projects (and this is what legacy transformation is, first and foremost) lies in this uncertainty. The reason behind this difficulty is that once the objections on the grounds of cost are addressed, then the ground shifts to questions about value (is it worth it?), and when that question is dealt with, the question of affordability (can we afford it?) is raised. After that, the discussion shifts to why it costs so much in the first place. And so the argument goes around and around.
The way out of the triangle is to identify the current position and then work systematically to achieve the right balance of arguments. At the "cost" apex, the central discussion is about how much to spend, what the options are, and which choice is the right one. At the "value" apex on the other hand, the focus is on how worthwhile the investment is — at its most basic, the objection goes back to “if it ain't broke, then don't fix it”. Finally, at the affordability apex, a favorable value and cost balance has been established, and the objections center on priorities and the other opportunities to spend this money in the business.
|Location||Central Issue||Symptoms||Way forward|
|Cost apex||Choice of solution||
A common thread running through the previous discussion has been the cost of doing nothing. Much of the cost of today's legacy is hidden in what might be called the "iceberg effect." On the surface, the only costs are the maintenance resource. But the real costs of a legacy application include equipment, software, personnel (operations, problem handling, liaison, maintenance, training, recruiting, and so on), communications (tariffs, support), and facilities (space, moves and changes, power, cooling). Business costs may include disruption to the business, downtime, staff overtime, rework due to failures/errors, or negative customer impacts.
The reason this is important to bring out in the open is because the costs of the legacy investment proposal will be visible and could look inordinately expensive if compared with the visible costs of the today's application. It is necessary to compare like with like.
According to Gartner, 60 to 80 percent of the typical IT budget is spent on maintaining mainframe applications and the applications that run on them. If transforming legacy applications can dent this figure by even 10 percent, the impact will be substantial. Further reductions can come from consolidating servers and storage — Gartner have pointed out that hardware expenses account for 18 percent of the typical budget (and the same amount again for operations staff). Consolidation can lead to 20 percent reduction in these costs (due to the improvements in utilization and load sharing). Thus it is important to show what the comparable figure is for the existing legacy applications.
The text-book definition of cost refers to “both the measurable and hard-to-measure resources for making goods and delivering services… the full cost of any cost object… is the cost of resources used directly for that object plus a share of the cost of resources used in common in making all objects”2. In this instance the costs are being collected for the purpose of making a decision so the precise allocation of resources is secondary. Allocating operating costs in this way can be a complex exercise, but the idea is to show the scale of the costs involved, not to tease out every detail. The resource allocation needs to focus on the areas where costs would vary if the transformation project goes ahead.
At the risk of stating the obvious, it will be management judgement that ultimately comes into play, and the figures are there to inform this decision.
This issue of visible/invisible costs applies equally to other options than legacy transformation. A case in point is legacy replacement: For example, the analysts looking at other options may have asked an ERP supplier for a quotation to replace the legacy application(s). It is necessary to go beyond this quotation, to consider the other (often substantial) costs involved, which are often well above and beyond the supplier's quoted price to do the work. These are another form of "iceberg" effect, although the 80/20 rule is less likely to apply:
Transforming legacy applications is a task with both risks and rewards. It is easier to rely on what seem like stable applications and hope that they will be adequate to keep the business going, at least in the medium term. But these legacy applications are at the heart of today's operations and if they get too far out of step with business needs the impact will be substantial, and possibly catastrophic. The challenge for the CIO is to present the arguments for the legacy investment in the best possible light, but also to give management the full picture of these risks and rewards so that they can make a decision in full possession of the facts. Ultimately, legacy transformation is an "enabling" project, that allows other things to happen, but it has its own direct benefits as well.
The discussion so far has concentrated on transformation technology and projects. In this section the discussion turns to the supply-side and where project managers should look to buy transformation technology and associated services. The concept of a "legacy transformation value chain" is introduced as a way of understanding the roles of different players on the supply-side. This is a way of thinking about the role that each type of player takes on in the value chain — from this the potential purchaser can take a view of how to go about selecting suitable supply partners.
The work needed to deliver transformation has several dimensions — requirements definition, program translation, testing, change management, installation of new platforms and so on. As explained earlier, the current state of development of the market means that there are few situations where the purchaser can rely on finding one supplier to satisfy all these needs, and the specialized know-how needed usually makes it impractical to undertake the work in-house. The purchaser will be looking to buy in some (or all) of the products and services needed, but will not want to deal with too many suppliers, to avoid the risks of split responsibilities and confusion over who does what. So who should the purchaser deal with?
A good starting point for understanding the supply-side is the "legacy transformation value chain." The diagram overleaf shows such a legacy transformation value chain. This value chain shows the different type of player — the providers of platforms and enabling projects, providers of transformation products and services, and the consumers — the IS department and the end-user (and possibly outsourcers).
The categories are somewhat arbitrary but essentially they reflect the source of supplier revenues. Note that a single supplier may take on more than one role. For example, selling consulting days as well as software licenses.
The best way to categorize the suppliers is to consider their revenue drivers. Each player in the value chain has different revenue drivers. The players upstream in the value chain (to the left of the diagram) are looking for on-going revenues, from licenses and upgrades, hardware and operating systems, training and support, and so on. The mid-stream players are mainly involved in the transformation project itself, so they will achieve what are essentially one-off revenues. Outsourcers and the in-house IS department for their part are concerned with the cost of operations and what revenues they derive from end-users and the business.
The value chain has three implications for supply-side behavior:
The best strategy is to strike a balance between the project-oriented players (who will take care of the transformation), and the infrastructure suppliers and in-house staff who have to make the end-result work every day, and to ensure that the transformation expertise is at the heart of the solution delivery.
At the same time, the purchaser will want to avoid dealing directly with a multiplicity of vendors and the risk that suppliers may "pass the buck" when difficulties arise.
The best strategy for the purchaser will look like this:
Legacy applications serve a vital role in the landscape of commerce. The service they provide is critical to the day-to-day function of industry as a whole. For example to conceive of e-commerce solutions without exploiting legacy applications is to ignore the one key to making end-to-end e-commerce a reality.
The most common complaint about legacy applications is that they are too old, too inflexible and too outdated to add real value to the evolution of industry system solutions. While an ideal world would discard last year's technology each year to replace it with the latest in system design and functionality, reality and prudence prevent this. In the case of mission-critical business systems, this is neither practical nor prudent. It is necessary to rethink how the life-cycle of legacy applications is managed. This report has argued that transformation is feasible, is becoming easier, and with the appropriate choice of strategy and project management, presents a real alternative to replacement or rewriting applications from scratch.
Transformation cannot happen overnight because it involves the collaboration of so many constituents with so much valuable business information and functionality at stake. However the availability of tools and processes can greatly simplify and speed up the process and, as even a brief consideration of the factors will reveal, the alternatives can be even more costly and time-consuming and risky for the business.
What is the future outlook? Current vendors' strategies are narrowly focused and suppliers have a proprietary interest in furthering their own tools and services. Without more co-operative effort across the value chain, this situation will continue to slow down the wider acceptance of transformation as a legitimate and effective way of moving forward.
On the other hand, the demand for access to the benefits of the present development tools available for the Java and .NET environments may drive transformation forward and increase the demand for transformation products and services. These environments include code generators, intelligent editors with built-in wizards, logic-tracing tools and debugging facilities. Analysts can create design models that feed specifications into development products. These tools are part of integrated development environments that synchronize business models, specifications and program logic across the development cycle. Changed business rules in a design model are reflected in the system source code and a coding change will also be reflected within a design model. Such model-driven development and maintenance is a very effective and efficient way to evolve systems over the long term.
Another factor is the lack of skilled resources: There are too many critical legacy applications and too few skilled technicians to work on them. Transformation provides the opportunity to increase the effectiveness of the legacy programmers through better analysis, development, upgrade, debugging and componentization tools. It makes sense to migrate legacy applications to these environments rather than look for the tools to be retrofitted to legacy environments.
Putting these factors in perspective, legacy transformation is currently at the early adopter stage, but if the level of interest in J2EE and .NET keeps increasing and users are looking to take the benefits of these platforms, then the future of the technology looks positive.
Declan Good has many years of experience in maximizing value from technology investment. At one time head of computer planning and research at the Canadian Customs and Excise, his subsequent career has focused on technology investment and technical architecture projects in both Canada and the United Kingdom.
He joined management consultants Woods Gordon in Ottawa (now Ernst & Young) after leaving the Canadian Government, and later worked with IT strategy consultants Butler Cox in the United Kingdom. He has been an Independent Consultant since 1992. Declan has degrees in engineering from Carleton University, Ottawa, and University College, Dublin.