ACM TechNews

Conquering Complexity

Computer (12/07) Vol. 40, No. 12, P. 111; Holzmann, Gerard J.

Minor software bugs can conspire to facilitate major system failures when combined, writes Gerard J. Holzmann of the NASA/JPL Laboratory for Reliable Software. "The probability of any one specific combination of failures will be extremely low, but as experience shows, this is precisely what leads to major accidents," he notes. Holzmann points out that the addition of fault protection and redundancy, while reducing the severity of failures, also makes a system bigger and more complex, which can unintentionally broaden failure modes by introducing unplanned linkages between otherwise separate system elements. Reduction of minor software defects can be achieved in a number of ways, including adherence to stricter coding standards, such as the required usage of strong static source code analyzers on every software build. Another strategy is to boost the amount of decoupling between software components, thus separating independent system functionality to the maximum degree possible. Executing independent functions on physically distinct processors supplying only restricted interfaces between them is one of the most powerful decoupling tactics. Holzmann also suggests that defects can be contained through the use of memory protection to ensure that multiple threads of execution in a computer cannot corrupt each other's address space, while supplying more margin than is required for system operation is still another approach to defect containment. Redundancy in safety-critical code can be delivered through the use of multiple functionality layers, while the most commonly utilized strategy is defect detection that commences at the very beginning of the software development process.

http://www.computer.org/computer

---

© Copyright 2008 Information, Inc. This service may be reproduced for internal distribution.