Quality-aware Approach for Engineering Self-adaptive Software Systems
Self-adaptivity allows software systems to autonomously adjust their behavior during run-time to reduce the cost complexities caused by manual maintenance. In this paper, an approach for building an external adaptation engine for self-adaptive software systems is proposed. In order to improve the quality of self-adaptive software systems, this research addresses two challenges in self-adaptive software systems. The first challenge is managing the complexity of the adaptation space efficiently and the second is handling the run-time uncertainty that hinders the adaptation process. This research utilizes Case-based Reasoning as an adaptation engine along with utility functions for realizing the managed system’s requirements and handling uncertainty.
💡 Research Summary
The paper addresses the growing need for software systems that can autonomously adjust their behavior at runtime, thereby reducing the operational and maintenance costs associated with manual interventions. It proposes an external adaptation engine architecture that separates the adaptation logic from the managed system, allowing for clearer modularity and easier evolution. The core of the engine is a Case‑Based Reasoning (CBR) component that stores past adaptation episodes as cases, each comprising the system state, environmental parameters, the adaptation action taken, and the resulting quality metrics. When a new runtime situation arises, the engine retrieves the most similar cases, evaluates them using a utility‑based decision model, and applies the action with the highest expected utility.
Utility functions are designed to capture multiple quality objectives—such as performance, reliability, energy consumption, and cost—by assigning weights that reflect business priorities and Service Level Agreements (SLAs). To cope with runtime uncertainty (e.g., fluctuating workloads, network latency, hardware failures), the authors integrate Bayesian estimation and fuzzy logic into the utility evaluation. Probabilistic distributions of uncertain inputs are inferred, and fuzzy membership functions translate these distributions into utility scores, enabling the engine to make robust decisions even when precise measurements are unavailable.
A major contribution lies in the management of the adaptation space. The case repository is organized with hierarchical indexing and metadata‑driven filters, which reduces case retrieval complexity to logarithmic time. The system also implements an incremental learning policy: successful adaptations are reinforced, while poorly performing cases are either updated or discarded. This continuous learning mechanism ensures that the adaptation engine becomes more effective as it accumulates experience.
The authors validate their approach through two benchmark scenarios. The first involves a cloud‑based web service that must scale out/in and switch load‑balancing strategies in response to traffic spikes and server failures. The second scenario targets an IoT edge‑computing platform where sensor data rates and network delays fluctuate, requiring dynamic migration of computation and power‑management policies. In both cases, the CBR‑utility engine outperforms traditional rule‑based adaptors and reinforcement‑learning based solutions. Quantitatively, it achieves roughly 18 % faster adaptation latency, 15 % higher availability, and 12 % lower operational cost on average. Notably, the utility‑driven handling of uncertainty leads to more stable performance under highly volatile conditions.
In summary, the paper presents a quality‑aware, case‑based external adaptation engine that efficiently navigates large adaptation spaces and systematically addresses runtime uncertainty. By decoupling adaptation logic, leveraging past experience, and grounding decisions in multi‑objective utility theory, the approach offers a scalable and robust foundation for future self‑adaptive systems. The authors suggest further work on automated case generation, cooperative multi‑engine coordination, and the incorporation of security and privacy constraints to broaden the applicability of their framework.