Application of Hierarchical Deep Reinforcement Learning for Dynamic Task Scheduling in Multi-Region Cloud Environments

Eda, Sri Lakshmi Prasanna (2025) Application of Hierarchical Deep Reinforcement Learning for Dynamic Task Scheduling in Multi-Region Cloud Environments. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration Manual) Download (550kB) | Preview

Abstract

Multi-cloud ecosystems provide organizations with flexibility and cost optimization but come with remarkable issues in job scheduling. The traditional scheduling algorithms cannot efficiently manage resource heterogeneity, while the deep reinforcement learning (DRL) methods that have been applied in the cloud environment encounter the scalability issue what with the centralized architectures that suffer state-action space explosion as cloud providers grow. This work presents an innovative framework called Hierarchical Deep Reinforcement Learning (HDRL) that is integrated with differential privacy as a part of the scheduling structure. The framework follows a two-tier building block, wherein the local agents carry out the resource allocation that is specific to each provider adopting proximal policy optimization (PPO) and the global coordinator that performs cross-provider decisions optimization using deep neural networks. This implementation validates the framework through multi-region deployment on AWS using Google Cloud Traces driven workloads for training and validation, directly demonstrates the feasibility and applicability of the method. The assessment of the AWS multi-region deployment delivered 46.7% cost savings, 83.3% accuracy in allocation and ε=1.0 differential privacy assurances with sub-100ms inference latency for allocation decisions. Thus, this research contributes to closing the gap between theoretical DRL models and production-ready, multi-region cloud, privacy-preserving, DRL-based schedulers.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Makki, Ahmed 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 12:48
Last Modified:	31 Aug 2026 12:48
URI:	https://norma.ncirl.ie/id/eprint/9696

Actions (login required)

View Item