GAO says FCS threatened by software bloat

The centerpiece program of Army transformation, Future Combat Systems, could collapse under the weight of poor software design, according to the Government Accountability Office.

In two reports released March 10, GAO said the total source lines of code for the $200 billion program have reached 95.1 million, almost triple the original 2003 estimate.

“Poor size estimation is one of the main reasons that major software-intensive acquisition programs ultimately fail,” GAO said.

The Army’s FCS program office, however, said the program is nowhere close to failure. “Fielding the FCS Brigade Combat Team is the cornerstone of Army modernization,” Army spokesman Paul Mehney said. “It is a commitment, not an option.”

FCS is a family of manned light tanks, robots and sensors that would be able to find and shoot enemies before they attack. The program’s success hinges on development of a communications network that links all components — and the network is largely a software function. Although GAO credits the Army and FCS’ lead systems integrator, Boeing, with attempting to put in place sound software development practices, total lines of code nonetheless have swelled to about four times more than other software-intensive projects such as the Joint Strike Fighter.

Mehney said the increase is mostly because of FCS officials’ decision to incorporate commercial technology, mainly the Red Hat version of the Linux operating system. “By using proven and secure COTS software, the Army is saving costs by reducing development and evaluation time,” he said. A better measure of program health would be to count the effective source lines of code, which takes into account how much code has been reused or adapted, rather than the total lines, Mehney said. Effective line estimates have been relatively stable since 2005, he said; GAO estimates a 15 percent growth to 19.6 million lines. However, GAO noted that there are other problems with software development.

Developers have complained that requirements have been poorly defined, arrived late or were unstable. As a result, developers had to defer some functions to future builds or waive them altogether to keep pace with the existing schedule, GAO found. Boeing program spokesman John Morocco noted that the build plan calls for an incremental development approach, which “provides the ability to learn from each previous build, as well as adjust to any changes in technology or priorities.”

But taken in total, evidence presented by GAO points to a program in trouble, said one defense software consultant, speaking on condition of anonymity. “The government does hardware buying really well, but software — they’re just incapable of doing it,” the consultant said.

Incremental development, for instance, is a good idea. But for it to work, it requires turnarounds of weeks rather than years, the consultant said. “The problem with software is that new languages come up, and also the computer hardware that the software runs on changes so rapidly,” he said. A model of quicker software development spirals would be possible even within the constraints of today’s complex acquisition environment, the consultant said. “It just takes leadership and somebody knowing how to do it.”