Shaping the Future of Software
Software Savings for NASA
Among her many connections outside UK, Hayes has fostered an ongoing relationship with NASA. Her current project is a collaboration with Alex Dekhtyar, a UK assistant professor in computer science and a Russian national, whose background proved to be an initial stumbling block.
"After the shuttle crash, NASA wasn't sure about letting non-citizens look at their data," Hayes says. "It took us a long time to get an agreement in place, but they decided to let Alex work on this. However, they're not allowing any of our non-citizen students to look at the data."
Jane Hayes and Alex Dekhtyar won the "Most Daring Research" award from NASA in July 2003.
This project focused on improving "requirements tracing," which means making sure every line of code is fulfilling one of the goals set for the software from the beginning. And the project netted the "Most Daring Research" award from NASA on July 30 of this year. The award is dubbed "The Buzz," in honor of astronaut Buzz Aldrin, who pioneered the idea of manned orbital rendezvous. Hayes says, "They felt we were daring to compare our approach to the de facto standard, a tool called SuperTrace. Our work has advanced and now we are out-performing the standard.
"We measure two things: recall and precision. Recall looks at whether or not we are finding all the items we should, and precision looks at how many incorrect items we are finding. Our recall is much higher right now than that of SuperTrace, with slightly higher precision."
Last year Hayes completed a NASA project on fault-based analysis, which, in a nutshell, identifies cause and effect. "In software engineering we get an indication that something went wrong, and we call that a 'failure.' We call the causewhich isn't always easy to identifythe 'fault.'"
NASA gave Hayes 17,000 fault reports, going back 15 years. "They basically gave us everything in their fault database."
When you talk about the International Space Station, Hayes says, you have millions of lines of code and billions of dollars in hardware and software, with 13 countries working together. "Obviously this is a huge management challenge, but just looking at the software piece, it's not straightforward as to what they should do to achieve the highest quality software possible. I could list 57 different approaches off the top of my head."
Hayes's research shows NASA how to get the most bang for their buck. "The whole idea here was to build a taxonomy of faults. The first thing I did was ask, 'What are the types of faults you see occurring across all NASA systems?' My taxonomy had 18 categories, but I found 85 percent of the faults fell into three of those categories. That information in and of itself was of huge value to NASAnow they know where to focus their efforts."
This isn't sexy stuff, Hayes says, but it's important.
Choose a section: