A Proactive Self-Adaptation Approach Based on Ensemble Prediction for Service-Based Systems
Shenglong Xie,Lu Wang,Qingshan Li,Xiangtian Guo
DOI: https://doi.org/10.1142/s0218126624502347
2024-03-16
Journal of Circuits Systems and Computers
Abstract:Journal of Circuits, Systems and Computers, Ahead of Print. Service-based systems (SBSs) are a unique category of software systems that dynamically combine various third-party services at runtime to deliver complex and adaptive functionality. This dynamic composition introduces a high level of unpredictability and uncertainty, creating potential anomalies and exerting significant pressure on system maintenance. To tackle this challenge, the conventional approach involves employing prediction-based proactive self-adaptation. Despite the prevalence of existing approaches emphasizing prediction accuracy, the critical aspect of "earliness" in predictions is often overlooked. Striking a balance between early and accurate predictions is paramount in practice. In response, we propose Proactive Self-Adaptation based on Ensemble Prediction (PSA-EP) to effectively balance the trade-off between the prediction earliness and accuracy of in SBSs. At the heart of PSA-EP lies an ensemble prediction model built upon a deep neural network and an enhanced long short-term memory (DNN-ELSTM) architecture. PSA-EP is crafted to empower SBSs to adapt to the inherent unpredictability and instability, facilitating the achievement of their adaptation goals. This adaptation mechanism not only enables effective prediction and analysis of adaptation goal violations but also addresses the reliability and performance of service level agreements (SLAs) governing quality of service (QoS). We evaluated the performance of PSA-EP in a decentralized tele-assistance system using four key metrics, and the experimental results, examined from various perspectives, underscore its exceptional performance.
engineering, electrical & electronic,computer science, hardware & architecture