By Tod Newcombe Contributing Editor When the business world first embraced client/server computing, all attention was on accomplishing one goal: replacing the costly mainframe with a network of less expensive computers. But companies soon found that migrating numerous applications and lots of data off of legacy systems was a lot trickier than they realized. The lack of quality client/server software tools, problems with data migration and staff resistance to change raised the cost and lengthened the development time for replacing mainframes with client/server applications. As a result, many organizations slowed down, postponed or gave up on the move. Even Sun Microsystems Inc., a leading vendor in the client/server field, kept its mainframe going for several years longer than it anticipated. "Surveys that we've done show a slowdown in the rate of moving applications off of legacy environments," said Mark Hess, vice president and research director of transition strategies for GartnerGroup Inc., a market research firm. "It was never very high in the first place." According to Hess, just a couple of years ago, the rate at which organizations were moving applications off of legacy systems was in the 10 to 15 percent range. Today, it's no higher than 8 percent. Instead of replacing mainframes, organizations are migrating just one or two legacy applications to a distributed environment and adding new applications - such as imaging, geographic information systems or a versatile case management system - that rely on mainframe data for some of their functionality. Wisely, governments aren't migrating applications to save on computing costs, according to Michelle Walsh, vice president for G2 Research Inc. One of the hard lessons learned in the early days of client/server is that lower costs are more an illusion than reality. Instead, said Walsh, agencies migrating to open up access to information. "That's why the commercial sector has already done it. It's pretty hard to work for Merrill Lynch if you don't have access to what you need - information. But until recently, if you worked for the Department of Motor Vehicles, it really didn't matter."
WHY MIGRATE? Speed gives client/server computing a major advantage over legacy systems, such as mainframes and minicomputers. Its speed can be defined in two ways. First, client/server technology divides the processing work between the server where all the data manipulation chores are done, and the client PC, which handles all the tasks relating to the user interface. By dividing the labor, client/server uses computers more efficiently - resulting in faster data processing. Second, development times for new applications are considerably shorter than they are on mainframes and minicomputer systems. Client/server applications can be developed in months instead of the years it sometimes takes to develop a mainframe application. And prototyping that application can be done in weeks rather than months. Along with speed, client/server offers greater functionality than can be found on mainframes. Shorter development cycles and easy-to-use development tools have spawned considerable creativity in the client/server world. Imaging, geographic information systems and multimedia are just some of the examples of new, more functional applications and software tools that are proliferating in government thanks to client/server. Finally, client/server gives the end-users much more control, allowing them to solve their own business problems rather than rely on mainframe programmers to develop a solution for them. For some time, government offices, departments and even agencies have viewed mainframes as too big, inflexible, bureaucratic, slow to change and too expensive. Client/server addresses all of those disadvantages with flexible, user-controlled solutions.
DRAWBACKS TO CLIENT/SERVER The number of computer applications in the private sector has exploded with client/server, and it's beginning to happen in government as well. Virtually everyone agrees that client/server has led to some very creative approaches to solving problems in business and government. But that creativity has come at a price, literally and figuratively. According to Hess, the idea that client/server costs less than mainframe computing stems from the practice of comparing the cost of MIPS (millions of instructions per second) on a mainframe with MIPS on a low-cost workstation. "The costs on a workstations are always far less," said Hess, "but that's only a one-dimensional comparison." While mainframes have high capital costs for hardware and software, client/server costs more when it comes to labor. What makes client/server so labor-intensive is the extensive support users need in a distributed environment. Support includes training, system management, security and network administration, as well as other needs. Mainframes, with their dumb terminals and text-based interfaces, are less prone to problems at the user end. And software development on client/server - while faster than on mainframes - is more complex because it requires the development of two programs - one for the client PC and one for the server. "When you look at the total cost of ownership, client/server actually costs more," said Hess. Another reason why organizations are reconsidering mainframe replacement concerns the value that exists in legacy applications. In many cases, applications developed in a legacy environment tend to be quality solutions, argued Hess. "Applications developed over a longer time frame tend to be done within an organizational framework, have consistent methodologies, are developed with re-use in mind, last over time, have economies of scale and more consistency in data integrity and security," he said.
MIGRATION STRATEGIES While complete mainframe replacement has become an unlikely proposition - especially for large organizations - so too is the notion that client/server is risky to implement. Government data centers may argue for the consistency and durability of mainframe computing, but today's service departments and agencies need computing that is flexible, easy to improve, and has low development costs. In other words, they need client/server technology. According to Hess, governments have to figure out how to balance these conflicting needs if they want to avoid battles over which computing strategy is the right one. "The most common strategy revolves around maintaining and sustaining the legacy environment for a set of needs, but handling all new requirements, developments and applications on distributed platforms," he said. Hess also said that agencies should migrate applications that need to be reengineered to meet the changing needs of government. The important thing to remember about client/server, commented Hess, is that it's not a wholesale style of computing that fits in everywhere. "Where it does fit in well is with applications that have to do with data access, where it is easier for end-users to get at data that may be stored on the mainframe, a server or on the desktop computer," he said.
OHIO'S MIGRATION PATH One example of the migration strategy outlined by Hess is taking place at Ohio's Bureau of Employment Services (OBES). The bureau provides the state's citizens with unemployment compensation and job-search services and administers the state's unemployment tax. These services require 2,700 bureau employees in 76 field offices statewide and in Columbus tokeep track of more than 5.5 million workers, 220,000 businesses and $1 billion in annual taxes. Automation has consisted of a 15-year old Unisys mainframe that hosts large databases and transactional programs pertaining to tax administration, employer histories and information on the unemployed. However, information relating to job applications, job openings and tax bills has remained on paper documents - a total of 10 million, to be exact. Besides the immense storage and retrieval problems created by paper documents, the bureau found that its existing paper-driven processes severely restricted its ability to deliver more effective services such as faster, more accurate job matches and streamlined billing of unemployment taxes. To correct these problems, the bureau awarded a contract to Unisys Corp. and a software vendor to develop an information and imaging solution that not only converted paper into electronic documents for rapid storage and retrieval, but also improved workflow so that new services could be delivered more readily. The system uses a Unisys 6000/85 server running UNIX, scanners, optical disc jukeboxes and 85 PC workstations, all running on a Novell local area network (LAN). The imaging software that drives the entire system is InfoImage Folder from Unisys. A key factor in the project was integrating the imaging solution with a tax database on the existing mainframe. The LAN is linked to the mainframe via an Ethernet connection running TCP/IP. The system allows users to simultaneously view document images and account histories from the mainframe database on a workstation. "We wanted to migrate as much functionality as possible away from the mainframe to our tax administrators," explained Don Russell, director of the bureau's Unemployment Compensation Tax Division. While the addition of the imaging application did not involve a full migration of mainframe data, it did reduce the Tax Division's mainframe data processing requirements. "We've reduced our demand for programming by 50 to 60 percent," said Russell. He added that without the integration of imaging with the mainframe database, the functionality of the system would have been reduced significantly. According to Mary Carroll, director of the bureau's Information Technology Division, the imaging system has had a positive impact on mainframe data processing. First, the integration required a minimal amount of programming to accomplish, primarily some minor formatting changes so that the mainframe could accept commands from the Visual Basic toolkit for data file transfers. Second, the imaging system has reduced demands for mainframe data processing time and labor, a valuable advantage in these days of shrinking budgets and rising needs for data processing. Taking a long-term view of the situation, Carroll said that new applications - like the imaging - will increasingly become the norm. "There's been a lot of technology changes recently," she said, "that give users new access to information and better control over its use." At the same time, she added, processing on the mainframe will occur less and less.
WHO WILL MIGRATE? According to studies by GartnerGroup Inc., organizations most likely to replace their mainframe with a client/server strategy are those with installations that average 25 to 30 MIPS in size. "These are the folks who developed their applications years ago and haven't sustained their investments," said Hess. "These organizations often find they can go out today and buy application packages that do most or all of what they need." Organizations with mainframe installations that are 100 MIPS or larger are unlikely to have a replacement strategy. Like OBES and other government agencies, they tend to add new applications or reengineer older ones into a client/server strategy. But invariably, their mainframe and client/server environment coexist. According to G2's Walsh, states are issuing more RFPs that specifically say that the agency wants to maintain the mainframe while adding new applications involving LANs and client/server. "Both environments have a clear, legitimate role today. It's not one or the other," said Hess. "The key is balancing strategies."
----------Sidebar---------- Connecticut's Only Choice HARTFORD, Conn. - Sometimes the legacy system has to go. That's the decision the Connecticut's Workers' Compensation Commission made when it brought its case management operations into the 1990s using client/server technology. By replacing a series of independent legacy systems with a statewide, integrated case management system, the Commission has seen productivity improve 30 percent since May 1995. The commission - which paid out $600 million last year on 60,000 filed claims - adjudicates and processes claims filed by workers and administers the compensation taxes of their employers. With headquarters in Hartford and each of the eight district offices running independent computer systems, the commission had little chance of effectively sharing data, monitoring workloads and analyzing information on a statewide basis. Furthermore, each district office developed its own unique procedures for performing tasks and capturing data. "There was no accountability," recalled Marvin Smernoff, chief administrative officer for the commission. "Meanwhile, the district offices were not forwarding information we could count on, so there was no credibility. In the meantime our staff kept growing." With encouragement from the state Legislature, the commission put out bids for a new system and awarded a contract to TRECOM Business Systems Inc., which proposed replacing the legacy systems with an integrated series of applications running on a client/server platform. Since the legacy applications were inadequate for the commission's more demanding needs, TRECOM ended up converting the legacy data to run on the new system. Ten Hewlett-Packard servers run the case management applications on a Sybase database management system for 200 Compaq PCs. Each district office also has office automation software on a local area network running Novell's Netware. Wide area communications are transported over a frame relay network. The $5.5 million system also supports several new technologies, including electronic data interchange for transmission of injury reports to and from insurance carriers, and the future use of imaging and workflow. According to Smernoff, the new system has shrunk case backlogs from eight months to 60 days by reducing the repetitive tasks of case workers. It has also made the commission more proactive in terms of analyzing claims, such as problems with fraud. This advantage is expected to reduce some of the costs of Workers' Compensation to employers. "As a result, we're becoming more of a help than a hindrance to businesses," said Smernoff.
