Abstract
The use of digital tools in education accelerated during the pandemic, highlighting platforms like Moodle and intelligent systems for personalized teaching. Smart Education is driven by 5G networks, which offer high data rates, low latency, and reliable connectivity, essential for multimedia applications. In this context, this article presents a proof-of-concept based on network slicing to optimize resources in a Distance Learning (DL) environment using the Moodle platform. An xApp is used to monitor the RAN and allow dynamic load balancing based on the profiles of connected users.
Introduction
The digital transformation of education has made platforms like Moodle central to learning. However, these platforms often struggle with "one-size-fits-all" network connectivity. In a typical distance learning scenario, a student downloading a text file and a professor streaming a high-definition lecture compete for the same bandwidth. While 5G promises high speeds and low latency, the challenge lies in application-aware resource management. This paper addresses the gap between the application layer (user roles in Moodle) and the network layer (resource allocation in the 5G Radio Access Network).
Methodology
The proposed system utilizes an Open Radio Access Network (O-RAN) architecture to implement dynamic network slicing. The prototype, named Slice App, operates on a Near-Real-Time RAN Intelligent Controller (Near-RT RIC).
The testbed was constructed using an open-source stack: OpenAirInterface5G (OAI) for the 5G Core and RAN, and FlexRIC for the controller. The xApp monitors Moodle session data stored in a MySQL database to identify user roles (e.g., "Common" student vs. "Privileged" professor). To manage traffic within encapsulated GTP tunnels, the authors employed Deep Packet Inspection (DPI) via the Scapy library and the Linux kernel's tc (traffic control) tool for Quality of Service (QoS) enforcement.
Experimental Setup
Instead of static datasets, the study employed a live experimental testbed. Virtual machines (VMs) simulated the 5G environment, generating real-time traffic through YouTube streaming and file downloads via a Moodle instance. This setup allowed for the evaluation of the system's ability to respond to dynamic user behavior in real-time.
Results and Discussion
The evaluation focused on two primary scenarios to test the efficacy of the role-based slicing mechanism.
In Case 1, representing standard users, the system successfully applied a slice that limited both users to 100 kbit/s, preventing resource hogging. In Case 2, when an "admin" user logged in, the xApp dynamically increased their throughput to approximately 230 kbit/s while maintaining the 100 kbit/s cap for the standard user.
A key technical insight was the comparison between metrics from the Core (via Bmon) and the RAN (via KPM). The results indicated that RAN-level Key Performance Measurements (KPM) provide a more accurate representation of the end-user experience compared to core-level monitoring. However, the authors noted that actual throughput was slightly lower than the theoretical limits due to the computational overhead of DPI and the simulation environment.
Conclusion and Future Work
This proof-of-concept demonstrates the potential of O-RAN and network slicing to create "smart" educational networks. By prioritizing traffic based on application-level roles, institutions can ensure critical activities, like a professor's live lecture, are protected from network congestion.
Future work will address the computational overhead of Scapy-based DPI by exploring more efficient kernel-level solutions like eBPF (Extended Berkeley Packet Filter). Additionally, the integration of Machine Learning models could enable proactive network adjustments, further enhancing the scalability and responsiveness of the system for large-scale university deployments.
Summary updated