Improving Performance Control of Edge Microservices with Proactive Autoscaling
Managing resources in cloud-edge applications is a complex challenge. While placing services closer to end users reduces latency and improves performance, it also introduces strict resource constraints and highly variable workloads.
In this context, traditional autoscaling mechanisms, such as those used in Kubernetes, react only after changes in traffic have already occurred. This reactive approach can lead to inefficient resource usage and potential violations of Service Level Agreements (SLAs).
To overcome this limitation, SMARTY introduces a proactive autoscaling approach based on machine learning, capable of predicting system behavior and anticipating adjustments in microservice configurations.
The proposed approach is built on a MAPE-K architecture (Monitor, Analyze, Plan, Execute, Knowledge), enabling real-time metric collection, data-driven analysis and automated scaling decisions. Through predictive models, the system estimates both resource consumption and the risk of SLA violations across different configurations, simulating the impact of scaling decisions before they are applied.
Experimental results demonstrate improved SLA compliance compared to traditional autoscaling methods while maintaining efficient resource utilization, making the proposed solution particularly suitable for resource-constrained edge environments.
Follow SMARTY for updates:
LinkedIn: https://www.linkedin.com/company/smartyeu/?viewAsMember=true
Website: https://www.smarty-project.eu