Back to search

University of Patras

Structuring and machine learning methodologies for knowledge mining and applications on big data

Abstract

dc:description

Educational institutions and medium-sized companies often possess significant underutilized computing power in their desktop hardware, while lacking the budget for dedicated high-performance computing (HPC) or expensive public cloud services. To address this gap, we designed and implemented CommC, a system that transforms existing commodity hardware into a cost-effective private computing cluster. The system leverages a hybrid architecture combining the Proxmox hypervisor for virtual machines, a layered Kubernetes cluster for orchestrating containerized applications, and lightweight Linux Containers for performance-critical tasks. The primary contribution of this work is an economic and architectural model for sustainable, on-premises computing. Our analysis demonstrates that the CommC cluster reduces computational costs by a factor of 6x to 8x compared to running equivalent workloads on major public cloud providers. To support this infrastructure, we developed a novel suite of management tools, including CommC Node Toolkit and CommC K8s Deployer for automating the installation procedure, CommC Provisioner for intelligent resource scheduling, SparkQueue for simplified job submission and SparkReact for interactive model creation. The platform's versatility and performance were validated through the successful deployment of real-world scientific applications, including the DiscMycoVir bioinformatics pipeline, running experiments for research projects in the field of NLP and the development of SparkBN, a scalable Bayesian Network classifier. The CommC project provides a proven blueprint for repurposing existing hardware into a powerful, secure, and economically transformative computing environment, freeing institutions from the recurring costs and vendor lock-in of commercial cloud platforms.

Degree

thesis:*
Grantor dc:publisher
University of Patras
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Bompotas, Agorakis
  • Μπομπότας, Αγοράκης

Subjects

dc:subject × 20

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/59912

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Bompotas, Agorakis; Μπομπότας, Αγοράκης. Structuring and machine learning methodologies for knowledge mining and applications on big data. University of Patras, 2025. http://hdl.handle.net/10442/hedi/59912