Home // ICSEA 2017, The Twelfth International Conference on Software Engineering Advances // View article

A Reusable Adaptation Component Design for Learning-Based Self-Adaptive Systems

Authors:
Kishan Kumar Ganguly
Kazi Sakib

Keywords: Reusable Adaptation Component; Environment Feature; Application Feature; Design Pattern

Abstract:
In self-adaptive systems, according to the separation of concern principle, the adaptation logic and the business logic components should be kept apart for reusability. However, this promotes reuse of the whole adaptation component while reuse of its subcomponents and their classes can also be helpful. Existing techniques do not consider this. Moreover, existing approaches also do not consider application and environment factors together for a more accurate adaptation. In this paper, a learning-based adaptation component design has been proposed which supports these. Machine learning is used to express metrics that measure system goals, as a combination of application and environment attributes. These are used to select application components to turn on or off by solving an optimization problem, aimed at maximizing system goal conformance. Components are turned on or off using a customizable effector component. Design patterns are utilized for increasing the reusability of the adaptation subcomponents. The proposed method was validated using the popular Znn.com problem. The reusability and learning accuracy metrics used indicate that it performs well for both. The system was also put under high load for observing adaptation of response time. It was seen that adaptation occurred as soon as the response time was over a provided threshold.

Pages: 244 to 249

Copyright: Copyright (c) IARIA, 2017

Publication date: October 8, 2017

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-61208-590-6

Location: Athens, Greece

Dates: from October 8, 2017 to October 12, 2017