Legacy Systems Back in the Limelight

Corporations look to reuse and extend legacy systems in an effort to expand their business model to the Internet; the death of COBOL is (still) greatly exaggerated.

Throughout the 1990s, large IT organizations learned — often the hard way — that building new corporate information systems from scratch was more likely to cause problems than solve them. Legacy commercial systems, most written in COBOL for IBM mainframe systems or proprietary minicomputer systems, still worked well and employees knew how to use them. A complete overhaul of these applications could cost organizations millions of dollars in terms of development and training costs, while causing disruption among users and customers.

A decade ago, experts proclaimed that the mainframe was dead and that new applications had to be written in new languages — read C and later C++, Visual Basic, PowerBuilder, Gupta, Delphi and the like — for the emerging client/server model. Analysts proclaimed that the client/server model represented the wave of the future, and that the server was not a mainframe running applications written in COBOL or Pascal. At the same time, IBM's mainframe business declined steadily and minicomputer makers like Digital Equipment Corp., Wang Laboratories Inc. and Prime Computer Inc. were being acquired and their proprietary technologies left to fade away. Unix servers and Windows clients were said to be the future for corporate IT organizations.

Yet many organizations continued to run, and still do to this day, on applications written for mainframes. In many cases, the smarter companies decided to wait and see how the client/server revolution would fall into place. By the end of the 1990s, the client/server bubble had pretty much burst as organizations concentrated instead on resolving Y2K issues and extending their businesses to the rapidly emerging World Wide Web. A key lesson learned from the so-called Y2K crisis was that efforts to replace legacy applications with new, internally built ones were mostly futile due to time and financial restraints. Thus, legacy applications retained their mission-critical status at many large corporations. The shift was a boon to IBM, whose mainframe business rebounded through the end of the 1990s.

The need to continue utilizing legacy applications caused a late 1990s change in strategy for Cary, N.C.-based Relativity Technologies Inc., said founder and Chief Executive Vivek Wadhwa. The company was founded about two years ago to help organizations rewrite legacy systems in Java, but Relativity executives quickly learned that corporate IT managers "don't want to transform [legacy] systems into Java," said Wadhwa. Java is best for building new applications that can be linked to or integrated with legacy applications.

So Relativity quickly shifted gears and concentrated its efforts on developing tools to "componentize" legacy applications to enable these components to be integrated with Web-based systems. "You can't rewrite systems that have been built over two or three decades," Wadhwa said. "The dark secret is that people don't know how their systems work. The opportunity is to componentize systems and integrate them with the new systems."

Today, many corporate IT organizations are emerging from completed Y2K projects with application backlogs and orders from the corner office to quickly build systems that can take advantage of the Internet's potential to significantly boost marketing/sales efforts. And many have learned through experience that utilizing and extending legacy applications can save the company significant development and training costs. In many cases, building new, mission-critical applications to replace existing ones is seen as financial suicide. In other cases, there is no choice — new systems must be built or bought because legacy applications may incorporate so many patches or be so old that they cannot be extended to support the requirements of today's organizations.

"This company has 20 years invested in COBOL systems," said Clyde Todd, section chief in the Internet division of the Army Air Force Exchange Service (AAFES), a non-profit retailer contracted to operate department stores on U.S. Army and Air Force bases in the United States and abroad. "We changed the front end. We have to keep up with our competitors off the military bases, the Wal-Marts and K-Marts [discount department store chains]. We kept the business logic on the mainframe when that was best. Our systems have grown over the past 20 years, but the core remains the same, and they will continue to [stay the same]."

The AAFES experience is not unique, observers say. According to data compiled by Giga Information Group, a Cambridge, Mass.-based market researcher, 70% or more of the world's active business applications are written in COBOL; 16,000 large enterprises worldwide still use COBOL in some way; and more than 2 million programmers are trained in COBOL development. Giga consultants also note that corporations that just spent millions to renovate COBOL systems to meet year 2000 requirements cannot justify starting a costly application replacement project, at least for the next two or three years. Thus, COBOL systems can be expected to run a slew of corporations for years to come.

The AAFES and other organizations are taking advantage of some of the advances to COBOL that have emerged over the past few years with little fanfare in an era of over-hyped, new Internet technologies and tools. These extensions to COBOL support component-based development, offer multiplatform support, add object-oriented technologies and can be used to Web-enable deep-rooted COBOL applications.

Liz Barnett, an analyst at Giga, noted that a variety of technologies can be used to extend legacy applications into the new world — including Web-to-host access tools, multiple middleware technologies, some of the application server types and legacy mining tools. But she warns IT managers to be wary of false promises. Not all tools can extend or upgrade existing applications.

"Buyers have to make sure they are buying the right solution," Barnett said. "You don't want a solution that's looking for a problem." She said that while some former year 2000 specialists have successfully produced capable solutions for extending COBOL applications, the technologies of others cannot yet be used to adequately extend legacy systems. And in some cases, she said, new development is best.

Analysts also warn that renewing and extending legacy applications is very complex. The task requires skills that are not always readily available internally or externally; in addition, companies must now battle each other to hire from a dwindling supply of software developers with expertise in these new technologies. Analysts at Giga Group list several skills that IT personnel need in order to implement renewal projects, including expertise in legacy systems, component and Internet development models, as well as expertise in integrating legacy and new applications, and knowledge of the packaged applications used at a site. Many IT groups are turning to outside consultants for help, but with varying degrees of success.

For example, some consultants do not include corporate developers in projects, which leaves a client with the same lack of expertise that existed when the project began. Development managers should view a consulting job as an opportunity to train internal developers in new technologies. More importantly, perhaps, they should ensure that internal engineers share their legacy systems expertise with consultants. Analysts note that any project that includes both legacy and new technologies requires personnel that can offer expertise in both systems.

Theory vs. reality

Organizations are "looking at how to mine the knowledge, the information that exists in legacy applications today," said Kevin Ashworth, director of the e-solutions practice at Boston-based consulting firm Keane Inc., which has converted at least part of its successful Y2K consulting business to a legacy renewal operation.

"Many companies have millions of dollars invested in systems that were built in-house. A lot of them were replaced with [packaged] CRM and ERP systems because of year 2000 issues, but there are still plenty of legacy applications that were not touched," he said.

Ashworth said Keane consultants are often called in to help IT developers "understand what a [legacy] application does. Many of these things are largely undocumented. With the advent of the Web, and efforts to connect them to the external world, the challenge is to understand them. One of the things we're asked by the bigger companies is 'How do we leverage what we have without having to go inside and document them by hand?'"

Though the question of how to best utilize legacy systems may be obvious, the potential answers likely require significant engineering expertise and monies. And experts agree that abiding by that theory is no guarantee that a specific legacy application will be renovated successfully. But for the most part, there is less risk in revamping a legacy application than in writing a new one.

If the search for new project skills to extend legacy systems leads to consultants, Giga recommends that any decision be based on thorough evaluation. Options can include traditional mainframe and client/server consultants; former year 2000 vendors; large systems integrators like IBM, Andersen Consulting and Electronic Data Systems; and the newer e-business service providers. Each offers different skills that may only work with specific projects.

Though IBM now pushes Java rather than COBOL as the development language for middle-tier apps, veteran COBOL players Merant plc, Rockville, Md.; Acucorp, San Diego; and Fujitsu Software Corp. (a unit of Tokyo-based Fujitsu Ltd.), San Jose, Calif.; continue to upgrade mainstay COBOL tools with object-oriented and Web development capabilities.

Keane's Ashworth also noted that organizations are utilizing tools from Jacada Ltd., Atlanta, and Netron Inc., Toronto, to add graphical interfaces to extended mainframe applications. Such tools "take screen scraping to another level," he noted.

IBM and Computer Associates International Inc., Islandia, N.Y., continue to sell and update the VisualAge for COBOL and Realia Workbench COBOL tools, respectively; but both are for building and maintaining mainframe-only applications, say experts.

The Army Air Force Exchange Service uses Merant's NetExpress tool for its "first step in moving [mainframe COBOL applications] to the Web," said Todd of the retailers' fast-action support team. Service developers built a full e-commerce site where "the true heart of the system still resides on the mainframe," he said. Todd said the development project did not require outside consultants because the organization could utilize its longtime COBOL expertise.

AAFES initiated its online business about four years ago. It has grown at more than 200% a year, bringing in about $25 million in revenue during 1999, said Todd.

Gerry Cullen, director of special projects at Detroit Diesel-Allison British Columbia, Vancouver, a manufacturer and distributor of diesel engines, said his firm first looked to abandon COBOL when IT managers looked to upgrade its Wang VS10000-based COBOL applications in the early 1990s. "First, we looked at a number of packages, but none of them fit the bill, especially when you throw in the cross-border [United States and Canada] requirements and some specialized things we have to do," he said. "So we shelved the project for a couple of years" before deciding to extend the legacy COBOL applications from the Wang hardware to a Unix-based system from the Data General unit of EMC Corp., Hopkinton, Mass., using Acucorp's Acucobol toolset, explained Cullen.

Cullen said the timing allowed Detroit Diesel developers to complete Y2K work as part of the process. "We had the conversion up and running in December 1998, and the Y2K conversion was done by June 1999," he said. He estimates that the decision to continue using legacy applications saved the firm approximately $500,000 in training costs for the 120 users of the system at the firm, which employs 190 and had sales of about $50 million last year.

Remaining a mainstay

Merant predecessor Micro Focus (Merant was formed by the merger of COBOL tool supplier Micro Focus and Intersolv in 1998) shifted strategies several years ago, from rehosting COBOL applications on non-mainframe platforms to offering tools for building distributed COBOL appli-cations and integrating them in non-COBOL environments.

"We've taken the old Micro Focus tools and added new technologies so developers can approach a problem the way they want to," said Joe Nicholson, vice president of marketing in Merant's Micro Focus unit. "[Organizations] want to push applications out to the Web, but not touch the code. When no code is touched, you reduce the risk of the code not working."

Acucorp's Acucobol tool family is described by Giga Information Group as a COBOL-like 4GL that can provide developers with most of the benefits of a client/server 4GL. "We're moving people off the mainframe to open systems — open COBOL systems," said Acucorp co-founder, President and Chief Executive Pamela Coker. "The main benefits we offer are very low risk and modernization at the same time." The Fujitsu toolset, also built to extend COBOL systems to new, more modern platforms, is still sold mostly in the Far East. The tool is said to hold a 40% share of the Japanese market, though the U.S unit is at times aggressive in its pursuit of business in North America.

Legacy systems and COBOL are mostly unknown by today's new breed of programmer, who concentrates on Java and Internet development. Yet these systems remain a mainstay in many organizations; experts advise the corporate world not to overlook the value of legacy code.