ircraft and his development staff the ground crew. "Rather than trying to build an entirely new airplane, we simply converted some of its key parts," he adds. That meant that First Image would stick with COBOL.
The approach
paid off
. At the start of the transition, First Image learned that its existing programs could run fine on Unix workstations and servers as long as programmers addressed some key technical issues. Then its latest acquisition--a division of NCR--imposed a breakneck deadline on the development staff to integrate NCR's additional 8000 mainframe programs into First Image's production system.
Every day that First Image would have spent rewriting those COBOL programs was a day the company wouldn't be recouping its acquisition costs. By staying with COBOL and rewriting only a fraction of the code, First Image installed its workstat
ion network and kept the business running efficiently enough to serve its existing customers and even take on some new business.
"A lot of people told us we should rewrite all our COBOL code in C," Altman recalls. "Two years ago, I might have agreed. Today, I realize that COBOL is a dynamic language in its own right with an enormous user base and a future. It's very open and transferable as an application language."
Staying Current
Although the NCR acquisition catapulted First Image's downsizing efforts, the company had already begun to shut down its mainframes. Altman wanted to pave the way for emerging technologies, such as imaging, document processing, database retrieval, and archiving, all of which were strongest on Unix and client/server platforms and stagnant on mainframe platforms.
These new technologies play directly into First Image's core business. Headquartered in Atlanta, Georgia, First Image helps other large corporations manage, organize, and distribute the
ir information more efficiently. "Most companies create more information than they can deal with effectively, making seemingly simple tasks, such as formatting and printing monthly account statements, an immense chore," Altman says. "Our specialty is organizing, manufacturing, and distributing information in the right format, where it is needed and when it is needed."
First Image's services include data acquisition and conversion, imaging, document processing, database retrieval, archiving, and distribution. For example, the company creates and distributes monthly account statements to a multinational customer base for one large mutual-fund group. And for the Ellis Island immigration authorities in New York, First Image regularly organizes immense databases of information onto easy-to-read CD-ROMs.
"We are a clearinghouse for people who want to transform electronic documents into any human-readable form or transform human-readable documents into any electronic form," Altman explains. In four yea
rs, First Image's annual revenues grew from $80 million to $250 million. Today the firm operates 90 sites in the U.S. and claims annual sales of approximately $275 million.
As the company expanded, it acquired several other companies, each with unique information systems. When First Image's managers decided to downsize, they knew the changeover wasn't going to be easy, but the motivation was clear. "To continue to grow, we had to establish a common [hardware and software] topology," Altman says. "We wanted a common, low-cost hardware platform to take advantage of the downward spiral in the cost of CPU power and peripherals. There was just this small matter of 15,000 old COBOL programs," he adds.
Glass-House Menagerie
First, the technical staff at First Image had to determine the best microprocessor-based platforms for COBOL. Many programmers consider COBOL a mainframe language, but First Image discovered that there are COBOL compilers on just about every type of computer, inclu
ding Unix workstations and PCs. "What runs on an IBM 370 or 390 platform will, for the most part, run on a smaller platform," Altman notes. "The compilers are mature, with well-established standards."
Sun Microsystems' implementation of COBOL on its SparcStation platform and Sun's Solaris OS particularly impressed Altman. Some of First Image's largest customers also rely on this Sun gear, which simplified compatibility issues. These factors convinced First Image to purchase 15 Sun SparcStation 20 and SparcStation 1000 workstations running Solaris 2.3, which now are connected to a TCP/IP WAN.
First Image then evaluated several COBOL development and run-time environments for the Solaris platform. The company especially needed an environment that included strong code-conversion tools, since this would determine the success or failure of a mainframe-to-Unix porting effort. Altman and his team established several criteria for the COBOL environment, including the following:
-- The ability to m
ove programs without modification.
-- Overall system performance on the workstations that was equal to
or better than the existing mainframe performance.
-- A COBOL compiler for Solaris that could accommodate new
development and ongoing changes to the code.
Selecting the Tools
After evaluating several Unix-based COBOL development environments against these criteria, Altman and his team chose Micro Focus and its family of products: Micro Focus COBOL for Unix, Micro Focus Toolbox for the Sun platform, and Micro Focus COBOL Workbench for DOS and Windows. Altman says he perceives Micro Focus as a market leader, and Sun recommended the products as well. He also considered AcuCobol and a competing product from Computer Associates--CA-Realia--but did not formally evaluate either one.
First Image uses Micro Focus COBOL for Unix as the main development environment. Together with the Toolbox, it integrates a COBOL compiler with workstation-based development and testin
g tools. The Workbench includes a graphical development environment, GUI data management tools, and extensive COBOL syntax and behavior support. Programmers can use the software to develop applications that run on mainframes, workstations, minicomputers, and PCs.
Altman says the visual nature of these tools aids productivity because developers can have several active windows open simultaneously. This means developers can test in one window, edit in another, compile in a third, and move from one to the next by pointing and clicking.
First Image's developers particularly like
Micro Focus's Animator
component, which lets them visually monitor the execution of a COBOL program for testing and troubleshooting. They can watch programs run and see potential problems without impacting other applications and processes on the production platform.
Before committing to Micro Focus, First Image tested it on some small COBOL applications. "Not only did the code convert from platform
to platform very easily, but we found the processor speed to be equal to or better than that of a large IBM mainframe running the same program," Altman says.
The real challenges and learning curve came as mainframe programmers strove to become proficient with Unix and its different editors, OS functions, ways of addressing system memory, and so forth. For a couple of weeks, First Image conducted formal training sessions in conjunction with Micro Focus; then it set up a core competency group to answer questions for the rest of the staff. "For the most part, we dove in and learned as we went along," Altman explains.
When developers were performing a few final tests with the new tools, an unforeseen event spurred them to action: the acquisition of the Data Copy Division of NCR. First Image had the option to either purchase the NCR division, along with its mainframe computers, or leave the mainframes out of the deal and simply purchase the mainframe code. It chose the code.
"We went from zer
o to 900 miles per hour in two days," Altman notes. "Suddenly, we had both the funding and the motivation to begin a large-scale mainframe-to-Unix port. We had to complete the code-conversion project before we could reap any real business benefit from the NCR deal. So we had to move fast."
And move fast they did, converting 8000 COBOL mainframe programs to the Solaris platform in just four months. "It was like walking through a dark room with foreign furniture," Altman recalls. "We knew nothing about the [former NCR] programs, we had no experience with the new computer gear, and we were using a COBOL compiler that we had little practical experience with."
How did the company make the daunting transition happen? Terry Wade, chief financial officer at First Image, says the philosophy that kept the company from getting into trouble as it carried out the downsizing project was a simple one: Do not try to change everything at once.
"Your costs can spiral out of control in a downsizing effort i
f you try to change too much too fast or expect all your mainframe programmers to suddenly become proficient with the C language and personal computers," Wade says. "In the early days of data processing, the computer was the most expensive resource. Now the people are the most expensive resource, especially the teams of programmers needed to write and rewrite software applications."
The project had its share of glitches. The first was a management misstep: Top-level managers lost sight of the project's focus, and the development staff had to remind them that the COBOL-conversion effort was just as important as any new development activities, because the former NCR programs were revenue-generating systems. "In [management's] view, the people devoted to swapping out the engine on the plane could have been devoted to building a new plane," Altman explains.
Technical problems centered around getting information off tapes and putting in place a consistent way to convert EBCDIC code to ASCII code. The
I/O capabilities of the Micro Focus products weren't designed for the variety of tape formats that First Image encountered. Altman's team had to do some programming to get the I/O routines off the tapes.
In hindsight, the technical staff's bold maneuver was equivalent to changing the engine on an airplane that's in flight, says Wade. "Since this code was the key ingredient to our revenue base, we couldn't afford to lose a single customer during the conversion," he notes. But the risk paid off. "With the mainframes out of the picture, we estimate an annual savings in computer operations and maintenance of $400,000," he adds. "That translates into an ROI of a little more than three years."
First Image kept costs down for the $1.5 million downsizing project in part because it didn't have to add additional staff. Also, the 10 Micro Focus COBOL Workbench licenses, at $3000 each, represented only about 2 percent of the total project cost.
To obtain a foothold on the new platform, First Image m
igrated large portions of the code base intact. The company rewrote only what had to be rewritten for performance or reusability reasons. Altman estimates that this amounted to only about 5 percent to 10 percent of the code. The small rewriting demand was one reason why the company didn't lose any existing business during the NCR conversion--and evn added some new business.
Smooth Landing
As the dust started to settle following the first wave of the program-conversion process, Altman and his colleagues devised a careful migration strategy to ensure a smooth landing. For example, all procedural code and applications programs remain written in COBOL "because of our investment and skill base," he explains. "Programmers come and go, but COBOL is verbose--almost self-documenting--making it easier for one developer to take up where another developer leaves off."
Perhaps more important, there's a huge reservoir of people who know COBOL, and these developers are often accustomed to pro
duction requirements. "COBOL developers know how to write programs that work every single time," Altman says. "Our C programmers have written some very sophisticated applications, but not one of them has written a system to deliver payroll, or a general ledger, or a core-business process."
Once it converted the NCR programs, First Image turned to its own mainframe programs at five different sites. In less than a year, the staff ported all the 15,000 programs from eight mainframe computers.
Staying Power
COBOL may be First Image's strategic choice for legacy applications, but when it comes to new development, the company's use of COBOL is strictly tactical. The company carries out new development with a combination of tools on PC and Unix platforms. For example, it used the Oracle 7 relational database and Sybase's PowerBuilder development tools to build a new order-entry system "simply because that made the best sense for this particular application," according to Altman.
While Oracle 7 can handle First Image's large production database requirements, the company chose PowerBuilder for its rapid application development (RAD) capabilities, such as live prototyping. "The trick is using the right tool for each project," Altman says. "When we need a batch process to convert images from machine-readable to human-readable form, we use Micro Focus COBOL. Our developers are striving to achieve a high degree of reusability so that we don't have to rewrite the code for each customer. We've begun to implement better inventory management and code management practices, and we are considering Object COBOL for upcoming tasks," he adds.
Some companies choose to take an all-or-nothing approach to downsizing and client/server challenges, Altman concludes. They want to adopt brand-new tools and simply throw away all the old ones. "But don't throw away your hammer just because you think you have a good screwdriver," he says. "For us, sticking with COBOL was a wise decision. It remains one
of our core competencies."
PRODUCT INFORMATION
Micro Focus, Inc.
Palo Alto, CA
Phone: (800) 872-6265 or (415) 856-4161
Fax: (415) 856-3724
Internet:
http://www.mfltd.co.uk
Circle 1003 on Inquiry Card.
When COBOL Is Cool
