SDN-based Network Slicing

This repository showcases the implementation of network slicing in an SDN to enable the isolation of network resources. The goal of this example is to show that different requirements can be fulfilled on a shared physical infrastructure by using network slicing.

Architecture Overview

The architecture employs a multi-hop topology for two emulation applications/tasks. The network comprises four hosts (h1, h2, h3, h4) and five switches (s1, s2, s3, s4, s5). See the picture below for visualization:

The network should be isolated into two slices for two services:

1. Video traffic slice: The video traffic (UDP port 9999) is able to gain max. 10Mbps bandwidth. The right of video traffic to use it is not affected by other traffics.
2. Other traffic slice: All traffics, which are not type UDP port 9999, can use any path of the network, but must not affect video traffic.

This repository contains the following files:

1. network.py: Script to build a network with four hosts and five switches, bandwidth is 1Mbps and 10Mbps.

2. topology_slicing.py: Application to isolate the network topology into upper slice (h1 -> s1 -> s2 -> s5 -> s4 -> h3, 10Mbps) and lower slice (h2 -> s1 -> s3 -> s4 -> h4, 1Mbps).

3. service_slicing.py: Application to isolate the service traffics into video traffic (UDP port 9999) obtaining 10Mbps and non-video traffic (the remaining services) obtaining 1Mbps.

Getting Started

Prerequisites

Make sure to have installed:

• Vagrant (>= v2.2.5)
• Virtualbox (>= v6.0)

Install then ComNetsEmu following these instructions. Additional documentation can be found here. ComNetsEmu is recommended to run in a VM with 2 vCPUs and 2GB RAM.

Installation

To launch ComNetsEmu, navigate to the ComNetsEmu directory and execute the following command:

  vagrant up comnetsemu

Once the VM when is up and running (indicated by the ComNetsEmu banner on the screen), SSH into the VM:

  vagrant ssh comnetsemu

Move to the ./app directory and clone the repository into a new local directory:

  git clone https://github.com/Faxatos/SDN-based-network-slicing.git

To halt the VM when done, utilize the following command:

  vagrant halt

Usage

You can simply run the emulation with following commands in the new local directory.

1. Enable Ryu controller to load the application and to run in background:

   $ ryu-manager topology_slicing.py &

2. Start the network with Mininet:

   $ sudo python3 network.py

Please stop the running Ryu controller before starting a new Ryu controller. For example, type htop in the terminal to show all running processes, press the key F4 to look for the process ryu-manager, then press the key F9 to stop the process, with the key F10 to quite htop.

How to Verify

There are three modes to verify the slice:

1. ping mode: verifying connectivity, e.g.

   mininet> pingall
   *** Ping: testing ping reachability
   h1 -> X h3 X 
   h2 -> X X h4 
   h3 -> h1 X X 
   h4 -> X h2 X 
   *** Results: 66% dropped (4/12 received)

2. iperf mode: verifying bandwidth, e.g.

   mininet> iperf h1 h3
   *** Iperf: testing TCP bandwidth between h1 and h3 
   *** Results: ['9.50 Mbits/sec', '9.98 Mbits/sec']
   mininet> iperf h2 h4
   *** Iperf: testing TCP bandwidth between h2 and h4 
   *** Results: ['958 Kbits/sec', '1.32 Mbits/sec']

3. client mode: verifying flows on each switch, e.g.

   mininet> sh ovs-ofctl dump-flows s1

Service chain

The inclusion of the s5 switch serves to implement a service chain, allowing the addition of network functionalities (such as firewalls) by modifying the switch flows. This can be achieved either with code or through the CLI. For simplicity, we'll utilize CLI using the ovs-ofctl commands, compatible with any OpenFlow switch.

Here are some useful commands for deleting and adding switch flows:

sh ovs-ofctl del-flows <switch> in_port=<port>
sh ovs-ofctl del-flows <switch> <protocol>
sh ovs-ofctl add-flow <switch> in_port=<port>,priority=<priority>,actions=<action>
sh ovs-ofctl add-flow <switch> <protocol>,priority=<priority>,actions=<action>

Here's an example of how this can be applied within the described architecture:

sh ovs-ofctl add-flow s5 icmp,priority=2,actions=drop

By adding this rule, all ICMP packets passing through s5 will be dropped. You can test this using the ping command.

Acknowledgments

This example is based on one of the comnetsemu applications.

Released under the MIT License.

License

Distributed under the MIT License. See LICENSE.txt for more information.