Ashfaaq Farzaan

SDN-based Load Balancer for SOHO Environments 🌐⚖️

Project Overview

This research project focuses on the development of a Software Defined Network (SDN) based load balancing algorithm tailored to enhance network performance, particularly in Small Office/Home Office (SOHO) environments. Traditional load balancers can be expensive and inflexible. This project proposes a cost-effective and scalable solution by implementing the load balancer as an application on an SDN controller.

The core idea is to distribute client requests to servers based on the least bandwidth utilized by a server at a given time. This approach aims to optimize resource utilization and improve response times. The system is designed to gather server bandwidth information over a separate wireless network to avoid congesting the primary data links.

🧠 What It Does

  • Distributes Network Traffic: Implements a load balancing algorithm on an SDN controller (POX) to intelligently distribute incoming client requests across a cluster of servers.
  • Optimizes Server Selection: Selects the most suitable server by identifying the one with the least current bandwidth usage.
  • Dynamic Bandwidth Monitoring: Servers periodically report their bandwidth utilization to the SDN controller.
  • Reduces Management Traffic Overhead: Proposes using a separate wireless network for transmitting server bandwidth data to the controller, preventing interference with client-server data traffic.
  • Comparative Analysis: The proposed algorithm’s performance (response time) is evaluated against traditional random and round-robin load balancing algorithms.
  • Versatile Implementation: Tested in both an emulated environment (Mininet) and a physical setup using Raspberry Pi devices.

🧰 Technical Highlights

  • Controller & Language: The load balancing application is developed as a module for the POX controller using Python.
  • Core Algorithm:
    • The _pick_server function within the POX module reads bandwidth values from a file.
    • It parses these values, sorts them to find the server with the least bandwidth, and directs traffic accordingly.
    • The MemoryEntry class helps in remembering flow states.
  • Bandwidth Data Collection:
    • A shell script designed to run on servers listed in /home/pi/serverlist.
    • Uses ssh to connect to each server and bwm-ng to capture network interface (eth0) bandwidth (specifically the ‘out’ rate).
    • Appends the collected bandwidth data to a central file (/home/pi/values) for the controller to access.
  • SDN Architecture:
    • Decouples the control plane (POX controller) from the data plane (OpenFlow-enabled switches).
    • Clients and servers connect to an OpenFlow switch (e.g., Open vSwitch).
    • The controller manages flow table entries in the switch to direct traffic.
  • Testing Environments:
    • Mininet Emulator: Used for creating virtual networks with multiple hosts and switches for initial testing and development.
    • Raspberry Pi: Implemented on Raspberry Pi devices, with one acting as an OpenFlow switch (running Open vSwitch) and others as hosts/servers, demonstrating real-world applicability.
  • Network Setup for Bandwidth Reporting: The design incorporates a separate wireless network for servers to transmit their bandwidth information to the POX controller, minimizing impact on the data network.
  • Performance Metrics: Primarily focused on response time under varying numbers of requests and server cluster sizes, compared against random and round-robin algorithms.

🔗 Explore the Project