for the most part, even if they're not exactly stuffed with 21st-century functionality.

At a certain point, old systems' idiosyncrasies -- relying on batch processing, the limited (or nonexistent) ability to run ad hoc reports, new program requirements that translate into costly and labor-intensive programming changes -- simply outweigh their usefulness.

This is when IT departments find themselves painted in the corner.

If they had their druthers, agencies would replace the systems. Who doesn't want new technology? The decision to migrate or not to migrate comes down to comparing the cost of supporting aging IT systems with the expense of implementing new technology.

Adding a cloud of uncertainty is that moving to modern systems is clearly a process fraught with danger: What if implementation goes awry? What if costs shoot through the roof? What if users resist the new system? What if you bet wrong on a particular type of technology? What if things don't work?

History Lesson

Perhaps the most inglorious attempt to scrap an aging IT system and move to a new one happened in California. The state's Department of Motor Vehicles (DMV) relied on archaic mainframe environments and databases to process driver's license and vehicle registration transactions. Further complicating the situation was the fact that the databases were written into the operating system, which made for fast searches but caused headaches when attempting to program changes into the systems.

The state reached a decision in the mid-1980s, after undertaking a feasibility study, to replace the system -- a process that dragged on until late 1993, when the project was finally killed by then-DMV director Frank Zolin.

When he became director of the DMV, Zolin inherited the ambitious systems-replacement project, which sought to merge the driver's license and vehicle registration databases into one gigantic database.

Zolin eventually resigned, along with two other key state officials, because of the failed project's fallout. The mess generated widespread anger and frustration as a slew of newspaper stories detailed the failure's costs (anywhere from $43 million to $50 million, depending on the newspaper); alleged malfeasance on the part of state officials; and strained relationships between the state and the vendor chosen to implement the new system.

"The status of the project was quite different than my preliminary briefings," Zolin said in discussing his initial surprise when reviewing the project. "I became director in March of 1991. I probably didn't realize the project was in trouble until that summer -- July or maybe even August."

He said the hardware -- Tandem Cyclone mainframe computers with a proprietary architecture -- was already purchased when he took over. Though the selection of Tandem hardware later caused plenty of controversy when the project was in its death throes, Zolin said the hardware was never the problem.

"The problems all stemmed from the software and the inability to program our business practices properly," he said. "The bottom line is it was the software development that failed."

The original contractor handling software development continuously fell behind and missed benchmarks, he said, and the decision was reached before he took over as director to buy out that contract and transfer software development responsibility to DMV data processing staff.

Zolin said if he had not taken over direct control of the project when he took the job, he might never have known its true status.

"I would have probably never killed the project the way I did if I hadn't had that personal involvement and gained that knowledge," he said. "If I had just depended on reports, I wouldn't have done it.

"I didn't know the project was in trouble until the summer, four or five months after I got the job, but that doesn't mean people didn't tell

Shane Peterson  |  Associate Editor