Source-linked AI summary
Perpetual Assurances for Self-Adaptive Systems
Danny Weyns, Nelly Bencomo, Radu Calinescu, Javier Cámara, Carlo Ghezzi, Vincenzo Grassi, Lars Grunske, Paola Inverardi, Jean-Marc Jézéquel, Sam Malek, Raffaela Mirandola, Marco Mori, Giordano Tamburrelli
TL;DR
Self-adaptive systems must assure requirements despite uncertainty and change across their lifetimes. The paper develops a framework for perpetual assurances, covering challenges, requirements, mechanisms, and benchmarks, and presents a service-based exemplar for comparing approaches.
Problem
Method
The paper develops a framework for perpetual assurances, outlining challenges, solution requirements, assurance mechanisms, benchmark criteria, and a service-based exemplar.
Source: §1 · p. 2, §3 · p. 6, §7.3 · p. 22Results
The paper identifies challenges and requirements for perpetual assurances and presents a service-based exemplar researchers can use to assess and compare approaches.
Source: §1 · p. 2, §8 · pp. 27–28Takeaways & Limitations
The exemplar provides a basis for comparing solutions, pinpointing critical challenges, and demonstrating the effectiveness of mechanisms for self-adaptive systems.
Source: §8 · p. 28The required level of assurance depends on the requirements of the self-adaptive system under consideration.
Source: §5.1 · p. 12
Abstract
from arXiv · showhide
Providing assurances for self-adaptive systems is challenging. A primary underlying problem is uncertainty that may stem from a variety of different sources, ranging from incomplete knowledge to sensor noise and uncertain behavior of humans in the loop. Providing assurances that the self-adaptive system complies with its requirements calls for an enduring process spanning the whole lifetime of the system. In this process, humans and the system jointly derive and integrate new evidence and arguments, which we coined perpetual assurances for self-adaptive systems. In this paper, we provide a background framework and the foundation for perpetual assurances for self-adaptive systems. We elaborate on the concrete challenges of offering perpetual assurances, requirements for solutions, realization techniques and mechanisms to make solutions suitable. We also present benchmark criteria to compare solutions. We then present a concrete exemplar that researchers can use to assess and compare approaches for perpetual assurances for self-adaptation.
1 Introduction
Self-adaptive systems shift change activities from development to runtime, making enduring evidence of requirements compliance important despite uncertainty. The paper frames perpetual assurances and outlines its challenges, solution requirements, techniques, benchmarks, and exemplar.
1 Introduction
- Self-adaptation shifts change activities and responsibility to runtime, blurring the development/runtime boundary and challenging traditional software engineering processes.Business- and safety-critical systems need evidence of requirements compliance throughout their lifetime despite uncertainty.Source: §1 · p. 1
- Perpetual assurances combine human- and system-driven activities to provide new evidence that requirements are met throughout a system’s lifecycle.The paper presents challenges, requirements, realization techniques, benchmark criteria, and a case study for comparing approaches.Source: §1 · p. 2
2 Key Challenges for Perpetual Assurances
Uncertainty is the central challenge to assuring self-adaptive systems across their lifetimes. The section frames perpetual assurances as an enduring process that combines human- and system-driven evidence, then classifies uncertainty and describes its sources.
2 Key Challenges for Perpetual Assurances
- Perpetual assurances provide requirements-compliance evidence through an enduring process that combines human- and system-driven activities across the system’s lifetime.The process continuously derives and integrates new evidence and arguments to assure stakeholders despite uncertainty.Source: §2 · p. 2
- Uncertainty can concern input parameters, model structure, or context; it ranges from deterministic knowledge to total ignorance and may be epistemic or aleatory.Ignoring uncertainty can lead to unsupported claims about system validity or generalization beyond a bounded scope.Source: §2 · p. 3
- Uncertainty arises from system modeling and adaptation, goals, execution context, and human aspects, including incompleteness, model drift, sensing noise, and human behavior.Sources may overlap and compound, so assurance solutions and their evidence must be able to address composed uncertainty.Source: §2 · pp. 4–5
6 Benchmark Criteria for Perpetual Assurances
The section proposes benchmark criteria for comparing perpetual-assurance approaches across capabilities, evidence basis, assurance stringency, and performance. Together, the criteria cover functional and quality requirements.
6 Benchmark Criteria for Perpetual Assurances
- The framework benchmarks approaches across four aspects: assurance capabilities, evidence basis, stringency of assurances, and performance, covering functional and quality requirements.These criteria are intended to compare different approaches to perpetual assurances.Source: §6 · p. 15
6.1 Capabilities of Approaches to Provide Perpetual Assurances
- Capability criteria include adapting to changing requirements and system components, handling inaccurate or incomplete models, and balancing evidence utility against its time and resource costs.When uncertainty is partly resolved, approaches may also support runtime reevaluation; system changes may be automatic or involve people.Source: §6.1 · p. 15
- Capability benchmarks also consider changing user loads and preferences, as well as flat or composed adaptation alternatives that may pre-empt one another.Changing request load can affect a service’s assured availability, while adaptation may reconfigure the system to maintain the required level.Source: §6.1 · pp. 15–16
6.2 Basis of Assurance Benchmarking
- Evidence-basis benchmarks distinguish historical data, future projections, combined approaches, and human input, and call for assessing historical relevance or prediction accuracy.Historical evidence may not guarantee future behavior; predictive approaches face the challenge of constructing models.Source: §6.2 · pp. 16–17
6.3 Stringency of Perpetual Assurances
- Assurance stringency concerns the rationale and evidence required, which may vary by system purpose and by adaptation phase.A formal proof may suit planning in one system, while statistical data may suffice elsewhere; confidence intervals may suit monitoring but not execution.Source: §6.3 · p. 17
6.4 Performance of Approaches to Provide Perpetual Assurances
- Performance benchmarks cover timeliness, computational overhead, and complexity, including the scope of problems to which an approach applies.Timeliness links directly to requirement R8, while human evidence links to R5, R7, and R10; handling alternatives relates indirectly to requirements through adaptation support.Source: §6.4 · pp. 17–18