Managing and Ensuring the Integrity of Non-Functional Requirements in Critical Software Systems
Carl Chang1, Peter In2, Jane Cleland-Huang3, Yujia Ge1, Tae-Hyung Kim1, JinChun Xia1 Iowa State University1 Dept. of Computer Sci. chang@cs.iastate.edu Texas A& M University2 Dept. of Computer Sci. hohin@cs.tamu.edu DePaul University3 Dept. of Computer Sci. jhuang@cs.depaul.edu
Abstract
This paper outlines an innovative approach to maaging and ensuring the integrity of non-functional requirements in critical software systems. NORM, which is an acronym for Non-functional requirements Optimization and Regression Monitoring addresses the crucial need for optimizing and maintaining nonfunctional qualities within critical systems in order to ensure their long-term integrity. The proposed method embraces the reality of fuzziness and imprecision by integrating fuzzy logic into an innovative technique for negotiating, modeling, optimizing, and balancing systemwide non-functional requirements. An agent-based compliance checking technique is proposed for monitoring not only the external behavior of the system at runtime, but long-term compliance to non-functional requirements at the code and architectural level.
related to reliability, performance, security, usability, and safety. Furthermore these requirements must be carefully balanced in order to maximize the overall quality of the system. Development of such systems involves rigorous analysis of these non-functional attributes and methodical consideration for their implementation [2,3,4]. When a critical system does fail, it is often a result of failure to meet one or more of these quality requirements. In fact there are many documented examples of such severe and catastrophic failures [4,5,6,7]. Based upon actual data obtained from industry, an IDC study reported that one hour of down-time costs an average financial broker house $6.5 million, and an airline reservation system $90K, yet such events happen regularly [8]. Even more disastrously failure of missioncritical systems, such as the self-destruction of Ariane 5 forty seconds after launch, cost literally millions of dollars and resulted in the complete loss of the system. In this case, failure was caused by lack of conformance to a stated safety requirement that had been relaxed in order to meet a conflicting performance requirement [9,10,11]. Therefore there is an urgent need to develop new and improved methods for preventing these kinds of errors from occurring.

下一页