Toward Systematic Conveying of Architecture Design Knowledge for Self-Adaptive Systems

Abstract

With the increasing complexity and stringent requirements of modern large-scale distributed systems, well-structured representations of software design knowledge arise as a promising approach to keep delivering high quality products in a timely and cost-effective way. Although domain-specific architecture styles and reference architectures help in conveying such design knowledge, the lack of systematic and structured representations makes it hard to grasp design alternatives promptly and support well-informed trade-off analysis. This short paper presents DuSE-MT - a supporting tool for the DuSE approach to architectural design of self-adaptive systems. DuSE-MT implements: i) a generic met model or systematic representation of design spaces (DuSE), which enables automated architecture design and analysis, ii) a specific design space for the self-adaptive systems domain (SA:DuSE), iii) a set of metrics that capture quality attributes of resulting self-adaptive architectures, and iv) a multi-objective optimization approach to explicitly elicit trade-off decision by finding out a set of Pareto-optimal candidate architectures. Our approach has been evaluated in a case study involving self-adaptive cloud-based services.