A Proactive Hardware-Aware Pod Migration Approach for Real-Time Applications in Edge Clouds

Shahbaz, Majid (2025) A Proactive Hardware-Aware Pod Migration Approach for Real-Time Applications in Edge Clouds. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (4MB) | Preview

Abstract

Modern container orchestration platforms like Kubernetes have radically changed the way applications are deployed in hybrid and edge environments. While Kubernetes is very scalable and manages containers automatically with great efficiency, it does not natively support live pod migration which is an important feature for real time and latency-sensitive applications. In standard deployments, pods are terminated and restarted on other nodes during hardware failures or maintenance, causing unwanted service interruptions in systems that require continuous availability and low latency. In typical deployments, pods are terminated and restarted on different nodes during hardware failures or maintenance, causing unwanted service interruptions in systems that require continuous availability and low latency. To address these limitations and challenges, this research introduces a Proactive Hardware-Aware Pod Migration approach that designed to overcome these limitations by using deep reinforcement learning (DRL) to intelligently predict hardware resource bottlenecks and initiate live migration with the help of Checkpoint/Restore in User space (CRIU) ahead of failures. This approach considers hardware heterogeneity across edge nodes such as differences in CPU architecture and memory capacity – to determine optimal migration paths that maintain SLA compliance.

The experimental verification using actual deployment data from Amazon EKS environments shows considerable performance gains: 67.8% decrease in response time, 63.8% decrease in SLA violations, and 90% decrease in downtime. The system provides 100% migration success rate while ensuring proactive decision making in 85% of the migration cases in multi-tenant edge computing environments. Statistically significance tests validate all enhancements are statistically significant (p< 0.001) with large effect sizes, showing the DRL-based solution achieves an effective intelligent pod migration in edge clouds.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Samarawickrama, Yasantha UNSPECIFIED
Uncontrolled Keywords:	Kubernetes; Pod Migration; CRIU; Edge Computing; Deep Reinforcement Learning; Hardware-Aware Scheduling; Real-Time Applications
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > QA Mathematics > Electronic computers. Computer science > Computer Systems > Computers > Electronic data processing > Electronic data processing--Distributed processing > Edge computing T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science > Computer Systems > Computers > Electronic data processing > Electronic data processing--Distributed processing > Edge computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	31 Mar 2026 08:17
Last Modified:	31 Mar 2026 08:17
URI:	https://norma.ncirl.ie/id/eprint/9262

Actions (login required)

View Item