The Handbook of Engineering Self-Aware and Self-Expressive Systems

TL;DR

This paper provides comprehensive guidelines, architectural patterns, and a systematic methodology for designing self-aware and self-expressive computing systems, validated through case studies showing improved performance.

Contribution

It introduces a pattern-driven methodology for engineering self-aware systems, including architectural primitives and decision-making guidance, grounded in the EPiCS project.

Findings

The methodology effectively guides system design decisions.

Self-aware systems outperform non-self-aware counterparts.

Case studies validate the approach's practicality and benefits.

Abstract

When faced with the task of designing and implementing a new self-aware and self-expressive computing system, researchers and practitioners need a set of guidelines on how to use the concepts and foundations developed in the Engineering Proprioception in Computing Systems (EPiCS) project. This report provides such guidelines on how to design self-aware and self-expressive computing systems in a principled way. We have documented different categories of self-awareness and self-expression level using architectural patterns. We have also documented common architectural primitives, their possible candidate techniques and attributes for architecting self-aware and self-expressive systems. Drawing on the knowledge obtained from the previous investigations, we proposed a pattern driven methodology for engineering self-aware and self-expressive systems to assist in utilising the patterns and primitives during design. The methodology contains detailed guidance to make decisions with respect to the possible design alternatives, providing a systematic way to build self-aware and self-expressive systems. Then, we qualitatively and quantitatively evaluated the methodology using two case studies. The results reveal that our pattern driven methodology covers the main aspects of engineering self-aware and self-expressive systems, and that the resulted systems perform significantly better than the non-self-aware systems.