Optimizing GPU Resource Allocation and Scheduling using a Hybrid Scheduler

Kavali, Sai Nitish (2023) Optimizing GPU Resource Allocation and Scheduling using a Hybrid Scheduler. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (951kB) | Preview
Preview	PDF (Configuration Manual) Download (772kB) | Preview

Abstract

This research explores the evolving landscape of computational technology, focusing on the critical role of Graphics Processing Units (GPUs) in addressing complex computing problems. Initially designed for gaming and visualization, GPUs have transformed into powerful instruments, driving advancements in deep learning, scientific simulations, and video rendering. Despite their advantages, limited studies have delved into GPU scheduling in cloud environments, posing challenges for organizations managing GPU-intensive workloads. This research introduces a hybrid scheduler, combining FIFO and Availability-based methodologies for efficient GPU resource utilization. The scheduler’s objective is to maximize throughput and minimize resource wastage by orchestrating task execution across homogeneous GPUs. The study specifically investigates the application of this hybrid scheduler in training Convolutional Neural Network (CNN) algorithms, demonstrating its effectiveness in optimizing resource utilization and adapting to dynamic machine learning workloads. The research addresses the underutilization of GPUs by introducing allocation methodologies based on availability and FIFO, and other strategies, enhancing overall hardware resource utilization.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Heeney, Sean 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 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:	28 Mar 2025 15:09
Last Modified:	28 Mar 2025 15:09
URI:	https://norma.ncirl.ie/id/eprint/7352

Actions (login required)

View Item