By Edward Cone  |  Posted 2002-04-08 Print this article Print

The development process adopted by the FAA can boost efficiency and reliability—if it's deployed correctly.

! Spiral Development Snafus">

Gotcha! Lessons Learned the Hard Way About Spiral Development

Did you know that:

Data warehouse projects often fail for lack of a spiral approach
Bill Inmon, "the father of the data warehouse," preaches a spiral approach to building a corporate information repository. He argues it's impossible to reach this company-wide goal without first tackling smaller data-mart projects that help decision-making in individual departments. "If you just walk up and ask an end-user what he or she wants, they can hardly tell you," he explains in one article he has authored on the topic. But once shown an example of an application, users quickly start providing requirements and wish lists.

Inmon, though, also emphasizes the corporate benefits of a data warehouse will never be realized if these departmental projects are pursued independently. Instead, they should be used as stepping stones to providing a central place to store and analyze corporate data. The catch: don't try to skip steps.

If a project is low-risk, a spiral approach may not make sense
University of Southern California professor Barry Boehm says it's important to determine when it's smart to get off the spiral merry-go-round.

Take the known quantity. If you know the project's requirements in advance, you follow a well-understood architecture, the needs are not likely to change and stakeholders have all bought in, then it becomes riskier not to proceed down a direct development path, Boehm says. So if the requirements for the inventory management system you need to install at your widgets factory are essentially similar to those of the gadget factory inventory system you implemented last year, and you're confident there will be few surprises, then it makes sense to get off the spiral plan.

On the other hand, if the requirements are uncertain or in flux, it makes more sense to create a series of prototypes focused on clarifying project goals.

A spiral approach can focus stakeholders on their common interests in a project
At the FAA, this meant overcoming a history of feuds between its unions and management.

For example, a project to revamp the computer systems for the air traffic control centers that manage flights over the oceans required agreement on a common system to replace software that had been developed independently for facilities in Ronkonkoma, N.Y., Oakland, Calif., and Anchorage, Alaska. The FAA selected an Adacel Technologies system that was in use in New Zealand and asked Lockheed Martin to help with the implementation.

Since controlling air traffic over the oceans is really an international function, other aviation authorities were interested in taking it over if the U.S. failed to deliver an upgraded system. That made it easier to win over the union, says Nancy Graham, the product team leader at FAA. "They had the same interest I did in getting the best system at the most reasonable cost."

Graham involved the unions in selecting the base software project, then got them to sign an agreement covering the changes that would be required and a process for considering any further change requests. The project required the U.S.' three oceanic centers to "agree on things they've never had to agree on before," including not only software but procedures. New procedures, like less conservative standards for the minimum separation between planes, will also be devised to take advantage of the more precise navigation possible with the new system. Compared to those changes, the software issues are "nuances," Graham says.

Effective spiraling means even "prototypes" have to venture into the real world
This is one of the most difficult challenges at FAA, where bugs that would be acceptable elsewhere can have life-and-death consequences. Still, users provide richer feedback when they can use a system in production, rather than a test environment.

The FAA managed to put its User Request Evaluation Tool into limited deployment while it was still officially a research prototype. URET was classified as an advisory tool, meaning that it helps plan course changes and detect flight plan conflicts, but does not directly control traffic. With the Standard Terminal Automation Replacement System (STARS), which controls live traffic near an airport, the solution was to deploy an interim version, initially deployed to airports with little traffic. This Early Display Configuration still had to be formally tested and certified, but because it grafted the new user interface onto the existing system, it gave controllers early experience with minimal disruption to operations.

Senior Writer and author of the Know It All blog

Ed Cone has worked as a contributing editor at Wired, a staff writer at Forbes, a senior writer for Ziff Davis with Baseline and Interactive Week, and as a freelancer based in Paris and then North Carolina for a wide variety of magazines and papers including the International Herald Tribune, Texas Monthly, and Playboy. He writes an opinion column in his hometown paper, the Greensboro News & Record, and publishes the semi-popular EdCone.com weblog. He lives in North Carolina with his wife, Lisa, two kids, and a dog.

Submit a Comment

Loading Comments...
eWeek eWeek

Have the latest technology news and resources emailed to you everyday.