HAFLoop: An architecture for supporting Highly Adaptive Feedback Loops in self-adaptive systems
Artikel i vetenskaplig tidskrift, 2020

Most of the current self-adaptive systems (SASs) rely on static feedback loops such as the IBM's MAPE-K loop for managing their adaptation process. Static loops do not allow SASs to react to runtime events such as changing adaptation requirements or MAPE-K elements’ faults. In order to address this issue, some solutions have emerged for manually or automatically perform changes on SASs’ feedback loops. However, from the software engineering perspective, most of the proposals cannot be reused or extended by other SASs. In this paper, we present HAFLoop (Highly Adaptive Feedback control Loop), a generic architectural proposal that aims at easing and fastening the design and implementation of adaptive feedback loops in modern SASs. Our solution enables both structural and parameter adaptation of the loop elements. Moreover, it provides a highly modular design that allows SASs’ owners to support a variety of feedback loop settings from centralized to fully decentralized. In this work, HAFLoop has been implemented as a framework for Java-based systems and evaluated in two emerging software application domains: self-driving vehicles and IoT networks. Results demonstrate that our proposal easies and accelerates the development of adaptive feedback loops as well as how it could help to address some of the most relevant challenges of self-driving vehicles and IoT applications. Concretely, HAFLoop has demonstrated to improve SASs’ feedback loops’ runtime availability and operation.

Smart vehicle

IoT system

Self-improvement

Adaptive feedback loop

Self-adaptive system

Adaptive monitoring

Författare

Edith Zavala

Universitat Politecnica de Catalunya

X. Franch

Universitat Politecnica de Catalunya

Jordi Marco

Universitat Politecnica de Catalunya

Christian Berger

Chalmers, Data- och informationsteknik, Software Engineering, Software Engineering for People, Architecture, Requirements and Traceability

Future Generation Computer Systems

0167-739X (ISSN)

Vol. 105 607-630

Ämneskategorier

Inbäddad systemteknik

Datavetenskap (datalogi)

Datorsystem

DOI

10.1016/j.future.2019.12.026

Mer information

Senast uppdaterat

2020-04-03