Distilled Lifelong Self-Adaptation for Configurable Systems

TL;DR

This work tackles runtime self-adaptation of highly configurable software under time-varying workloads by introducing DLiSA, a lifelong planning framework that continuously leverages accumulated knowledge. It combines lifelong evolutionary planning with distilled knowledge seeding, guided by ranked workload similarity analysis, to dynamically and selectively seed past configurations when beneficial, formalized through $S_{t}^{t+1} = 1 - \mathcal{L}(\mathbf{D}_{t}^{t+1})/N_{pairs}$ and configuration weights $w_c = w_{c,r} + w_{c,t}$ where $w_{c,r} = O_c/H$ and $w_{c,t} = S_c/H$. The approach is evaluated on nine real-world configurable systems across 93 workload scenarios, showing up to $2.29\times$ improvements in efficacy and $2.22\times$ improvements in efficiency over state-of-the-art baselines. The findings demonstrate that distilling and selectively seeding past knowledge can significantly enhance both planning speed and the quality of adaptation under changing workloads, with implications for resilient, self-evolving software systems.

Abstract

Modern configurable systems provide tremendous opportunities for engineering future intelligent software systems. A key difficulty thereof is how to effectively self-adapt the configuration of a running system such that its performance (e.g., runtime and throughput) can be optimized under time-varying workloads. This unfortunately remains unaddressed in existing approaches as they either overlook the available past knowledge or rely on static exploitation of past knowledge without reasoning the usefulness of information when planning for self-adaptation. In this paper, we tackle this challenging problem by proposing DLiSA, a framework that self-adapts configurable systems. DLiSA comes with two properties: firstly, it supports lifelong planning, and thereby the planning process runs continuously throughout the lifetime of the system, allowing dynamic exploitation of the accumulated knowledge for rapid adaptation. Secondly, the planning for a newly emerged workload is boosted via distilled knowledge seeding, in which the knowledge is dynamically purified such that only useful past configurations are seeded when necessary, mitigating misleading information. Extensive experiments suggest that the proposed DLiSA significantly outperforms state-of-the-art approaches, demonstrating a performance improvement of up to 229% and a resource acceleration of up to 2.22x on generating promising adaptation configurations. All data and sources can be found at our repository: https://github.com/ideas-labo/dlisa.

Figures (5)

• Figure 1: Self-adaptation planning for configurable systems.
• Figure 2: An illustration of similarities and discrepancies in top 50 performing configurations across workloads. The same configurations are connected by dashed lines.
• Figure 3: DLiSA architecture for configurable systems.
• Figure 4: The sensitivity of DLiSA to different $\alpha$ values.
• Figure 5: Examples illustrating the workload similarity and seeded configurations with respect to the weights. In (b), the dotted lines highlight the selected ones for seeding.