The Mininet Approach to Network Research and Prototyping

Mininet was conceived as a way to make it easy to do network research using OpenFlow.

It is based on several guiding ideas and principles:

• It should be easy and fun to create OpenFlow networks
• Individual researchers and students should be able to easily create and experiment with networks on a single laptop
• Current laptops should be able to create very large experimental networks (with 1990s-class performance )
• Current clusters should be able to simulate enterprise-class (e.g. 25,000 nodes or more) networks
• Network topologies should be easy to create with simple parameters
• Python is (or should be!) simpler than XML for specifying network configurations
• A single network console that can control an entire network is better than hundreds of xterm or ssh sessions
• Full virtualization is overkill for most applications

Although Mininet was designed for OpenFlow networks, it also supports legacy IP networking, e.g. using Open vSwitch or an OpenFlow controller that supports standard Ethernet switching and IP routing.

Mininet: Squeezing a 1000-node Openflow Network onto a Laptop

Here is my original presentation to the OpenFlow Software Architecture and Implementation (SWAI) working group from 11/19/2009 (with an additional slide to illustrate the kernel datapath):

• "Mininet: Squeezing a 1000-node OpenFlow Network onto a Laptop"

-- BobLantz - 11 Feb 2010