Cost and Carbon Aware Reinforcement Learning Autoscaling for Multi-Cloud Kubernetes Spot Instances

Ghode, Vikrant Anil (2025) Cost and Carbon Aware Reinforcement Learning Autoscaling for Multi-Cloud Kubernetes Spot Instances. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

In a few public clouds, cloud-native microservices typically operate on managed Kubernetes. Elasticating the resources assists them in utilizing better, nevertheless, the cost of computation and emissions caused by the operations remains significant. Free but cancelable (such as Spot or Preemptible instances) are much cheaper yet may be terminated at any time and will have different effects on emissions depending on that area, so delivery is more difficult. This paper presents an auto-scaler which employs reinforcement learning to balance these three variables: cost, lifecycle carbon, and service reliability, through switching among clouds, regions, purchasing choices, and timing.

The auto-scaler monitors real-time demand, far tail-latency indicators, revocation risk, and predictions of carbon intensity and then decides what and how many machines to execute. It is benchmarked with genetic algorithms and threshold on typical workloads of microservices. The point is that a multi-objective RL policy may reduce costs and emission levels and retain additional SLO infractions within a low cap that provides an open and reusable strategy that users can apply to FinOps and greener cloud computing.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Gupta, Shaguna UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	31 Aug 2026 14:32
Last Modified:	31 Aug 2026 14:32
URI:	https://norma.ncirl.ie/id/eprint/9704

Actions (login required)

View Item