Architecture-Based Behavioral Adaptation with Generated Alternatives and Relaxed Constraints

作者：Chen, Bihuan^*; Peng, Xin; Liu, Yang; Song, Songzheng; Zheng, Jiahuan; Zhao, Wenyun

来源：IEEE Transactions on Services Computing, 2019, 12(1): 73-87.

DOI：10.1109/TSC.2016.2593459

• Behavioral adaptation
• relaxed constraints
• business value
• quantitative verification
• genetic algorithm

Software systems are increasingly required to autonomously adapt their architectural structures and/or behaviors to runtime environmental changes. However, existing architecture-based self-adaptation approaches mostly focus on structural adaptations within a predefined space of architectural alternatives (e.g., switching between two alternative services) while merely considering quality constraints (e.g., reliability and performance). In this paper, we propose a new architecture-based self-adaptation approach, which performs behavioral adaptations with automatically generated alternatives and supports relaxed functional constraints from the perspective of business value. Specifically, we propose a technique to automatically generate behavioral alternatives of a software system from the currently-employed architectural behavioral specification. We employ business value to comprehensively evaluate the behavioral alternatives while capturing the trade-offs among relaxed functional and quality constraints. We also introduce a genetic algorithm-based planning technique to efficiently search for the optimal (sometimes a near-optimal) behavioral alternative that can provide the best business value. The experimental study on an online order processing benchmark has shown promising results that the proposed approach can improve adaptation flexibility and business value with acceptable performance overhead.

• 出版日期2019-2
• 单位复旦大学; 南阳理工学院