Towards Self-Improving Hybrid Elasticity Control of Cloud-based Software Systems

TL;DR

This paper presents a novel hybrid elasticity control approach for cloud-based systems that combines infrastructure and software elasticity with proactive, reactive, and responsive decision-making to improve performance and cost-efficiency.

Contribution

It introduces a self-improving hybrid elasticity control method integrating multiple elasticity strategies at both infrastructure and software levels.

Findings

Enhanced resource management performance

Cost savings demonstrated in simulations

Effective handling of workload variations

Abstract

Elasticity is a form of self-adaptivity in cloud-based software systems that is typically restricted to the infrastructure layer and realized through auto-scaling. However, both reactive and proactive forms of infrastructure auto-scaling have limitations, when used separately as well as together. To address these limitations, we propose an approach for self-improving hybrid elasticity control that combines (a) infrastructure and software elasticity, and (b) proactive, reactive and responsive decision-making. At the infrastructure layer, resources are provisioned proactively based on one-step-ahead workload forecasts, and reactively, based on observed workload variations. At the software layer, features are activated or deactivated in response to transient, minor deviations from the predicted workload. The proposed approach can lead to better performance-aware and cost-effective resource management in cloud-based software systems. We validate our approach via a partial realization and simulation with real-world datasets.