By Tod Newcombe Contributing Editor The pressure to migrate aging computer systems is intensifying. A 1995 Deloitte & Touche survey of 400 public- and private-sector CIOs revealed that 75 percent still had legacy systems running mission-critical applications. And 45 percent of the respondents said their organizations would spend "significantly" more on client/server during the next two years. CIOs want client/server to get away from old problems as much as to pursue new opportunities. "Supporting a legacy system involves maintaining the application code and the data as much as it does maintaining the physical system on which that system is running," wrote Michael A. Gould in I.T. Magazine (April 1993). "Maintenance of the code, often COBOL or other equally inefficient languages, is an extremely costly process." Given all these reasons, why are Information Services (IS) Departments still so reluctant to migrate? One reason concerns the new computing architectures. Among states and localities, there is still uncertainty as to whether client/server systems are capable of handling critical applications in a cost-effective manner. UNIX systems are growing in popularity, but many IS departments lack the expertise needed to develop and maintain them.
REWRITING CODE Another reason concerns the applications in legacy systems. In order to use these new computing strategies, IS departments have to rewrite their legacy code - a daunting task, considering the size of the some of the applications. Why rewrite the code? Because without it, agencies lose a very precious asset. Old mainframe application codes contain the business rules that programmers captured during the original system development. These business rules are of great value, explained Gould, because of the "information that the code collects, accesses, and manages." Some government IS departments would rather avoid the migration issue altogether. Lack of resources and other factors means migration will have to wait. For them, the question is: 'how do we avoid rewriting our legacy code, preserve the older hardware and yet still move toward open systems?' Luckily, answers - in the form of special software tools - exist. Middleware is one software component that enables an organization to connect a variety of legacy databases together. Another emerging set of software tools are called "wrappers," which preserve the legacy code behind a software shell that can support new client/server applications. Both approaches extend the life of a legacy system. They don't preserve them, however, and eventually, government legacy systems will have to be rewritten.
CLOSING THE CONNECTIVITY GAP Sometimes, maintaining legacy systems means juggling a series of databases across a number of decentralized computing systems. Some databases are on legacy systems, others are newer, relational systems running on client/server platforms. If an agency wants to share information among these different systems, then the IS department needs in-depth expertise to run several databases so they can talk to each other. "We can eliminate the juggling," explained Al Sterling in describing his company's software tool for maintaining legacy systems. Sterling is the U.S. product marketing manager for Techgnosis Inc. of Boston, a developer of middleware. He described middleware as essentially a connectivity tool that eliminates a lot of the work that goes into sharing information between more than one database. "It's a problem for the '90s," said Sterling, who pointed out that the rush to develop new applications using less costly hardware and software during the '80s resulted in the problem of heterogeneous computing: isolated databases running on proprietary systems scattered throughout an organization. To solve the problem, firms such as Techgnosis have developed middleware products to close the connectivity gap between databases so users can more easily share information. "We provide accelerated communications across platforms," said Sterling. Techgnosis' key product is Sequel Link, which gives agencies saddled with legacy systems an avenue toward open systems technology. Its connectivity feature allows an agency to preserve legacy databases while connecting them to newer applications. "With middleware, an organization can pick its databases according to their merits rather than according to necessity," said Sterling. Good middleware tools shouldn't sacrifice performance for the sake of connectivity. According to Sterling, Sequel Link synchronizes and optimizes communications across a network. It eliminates the unnecessary packets of information that database management systems send out when they try to exchange information with another database. By cutting down on what Sterling calls "network chatter," Sequel Link can speed up a user's database query. Not surprisingly, middleware is most cost-effective for organizations that have a large number of disparate databases. "The more centralized an organization is, the less likely they need this type of middleware," said Sterling.
NAVIGATING WITHOUT STANDARDS Not to be outdone by third-party middleware products, developers of major database management systems have come up with their own sets of tools that help organizations preserve legacy systems while opening up connections to new generations of applications. The central problem, according to experts, has been standards - or the lack of them. "Old legacy systems run non-relational databases and today's applications can't talk to them," said Larry Lawrence, a sales manager for Informix Federal. Lawrence added that without standards, it becomes difficult for users to navigate their way to the information they need. "Trying to find information in different databases can be difficult given the variations in protocols, standards, error recognition and security," he said. To resolve that problem Informix has developed a suite of tools that allow users to navigate into a host of databases. One key tool, called the Enterprise Gateway, gives users standard query language (SQL) and remote procedure call (RPC) access to more than 60 relational and non-relational database management systems. Gateway is essentially middleware aimed at opening up and optimizing connectivity in a heterogeneous environment. According to Lawrence, middleware tools provide organizations with a transition phase between the old legacy systems and the new systems. The ultimate goal is to migrate applications away from systems that are costly to maintain toward today's less costly, more functional environments, such as UNIX and client/server. "But migration is not a trivial exercise," he said. The challenge any organization faces is cleaning the data and consolidating it. This is especially true if an agency plans to build a data warehouse: a centralized, integrated database of information designed to support decision-making and analysis. According to Lawrence, before data can be moved into a warehouse, an agency must get rid of duplicate data and then consolidate data that's been stored in different formats. "The point," said Lawrence, "is not to bring forward old data into one system, but to bring forward intelligence."
WRAPPING THE CODE Middleware isn't the only way to keep legacy systems alive. Some large organizations have turned to a technology known as wrapping, which enables newer applications to access legacy functions. Application wrappers surround a complete legacy system witha new software layer, which hides the legacy code and data from new client/server applications. Wrappers can save IS departments the time and money required to rewrite legacy code. They can also extend the life of mainframe systems that, in some cases, are better suited than client/server or UNIX to run certain applications, such as high-volume batch processing operations. "Wrapper software converts the character stream issued by the legacy system to a true API (application program interface)," wrote Paul Winsberg in Datamation (May 15, 1995). "New client/server applications can interact with the legacy system via the API." Wrappers can't be bought; they have to be built. But several products are available to help. The most advanced kind are object-oriented wrappers. Some of the vendors producing these kinds of wrappers include Micro Focus Inc. (Palo Alto, Calif.), Digitalk (Sunnyvale, Calif.) and Hewlett-Packard (Cupertino, Calif.). According to Winsberg, wrappers are similar to screen scrapers, "which help you build a graphical interface to legacy systems. However, a screen scraper converts the character stream to a GUI (graphical user interface), where an application wrapper converts it to an API. In the long run, wrappers are more useful because an API is more versatile than a user interface." Wrappers aren't perfect, of course. They can require a fair amount of tuning and debugging before they work. Some users of wrappers have complained about problems with performance and integrity. But despite the drawbacks, experts say that, in the right situation, wrappers can lengthen the inevitable transition of a legacy system to client/server.
