Solving Bufferbloat using Multilevel Feedback Queues (MLFQ)

Project Overview

This project implements a Multilevel Feedback Queue (MLFQ) algorithm to mitigate the bufferbloat problem in networking. The MLFQ prioritizes packet scheduling based on real-time requirements, ensuring low latency for critical applications like video conferencing, online gaming, and streaming services.

Overview

Bufferbloat occurs due to excessive packet queuing in oversized network buffers, leading to increased latency, jitter, and packet loss. This project presents an MLFQ-based scheduling mechanism that prioritizes packets dynamically to enhance Quality of Service (QoS) in networks.

Features

• Priority-based scheduling: Categorizes packets into different priority queues.
• Real-time adaptation: Uses a feedback mechanism to promote long-waiting packets.
• Latency Reduction: Minimizes delay for real-time applications.
• Efficient Bandwidth Utilization: Ensures fair distribution of bandwidth.
• Simulation of Network Traffic: Models packet scheduling in a simulated environment.

Installation

The research paper is not published yet, so the code is not available.

Usage

Modify the config.py file to customize simulation parameters like packet arrival rates, queue size, and priority levels. Then, execute main.py to analyze MLFQ performance.

Algorithm Details

1. Packet Classification

• Video conferencing packets (UDP, ports 16384-32767) → Highest Priority
• Online gaming packets (UDP, ports 3478-3480) → High Priority
• Streaming packets (TCP, keyword 'video') → Medium Priority
• File downloads (Other TCP/UDP traffic) → Lowest Priority

2. MLFQ Structure

• Level 0: Real-time critical packets (video calls, gaming)
• Level 1: Streaming packets
• Level N: Non-time-sensitive traffic (downloads, background processes)

3. Feedback Mechanism

• If a packet waits too long in a lower-priority queue, it gets promoted to a higher-priority queue to prevent starvation.

Results

The simulation evaluates MLFQ efficiency using the following metrics:

• Packet processing time
• Queue wait times
• Dropped packets due to congestion
• Improved latency for real-time applications

Future Work

• Implement machine learning for dynamic traffic classification.
• Optimize aging mechanisms to balance fairness and efficiency.
• Extend support for Software-Defined Networking (SDN) environments.

References

Relevant research papers and sources used in this project:

• J. Gettys and K. Nichols, "Bufferbloat: Dark buffers in the Internet," Communications of the ACM, 2012.
• K. Nichols and V. Jacobson, "Controlling queue delay," ACM, 2012.
• Various studies on SFQ, CoDel, and FQ-CoDel techniques.

---

Author

Developed by Balkishan