Magellan Compiler

Overview

This compiler has 2 functions:

Translate high-level network-wide SDN program to P4 programs for each device
Generate runtime program on Controller

Get Started

Here we get started by compiling and running the L2 network which supports Mac learning and STP, the program is as follows

from topology import shortestPath, spanningTree

global_mac_table = {} # mac -> the 'h->e' port ('00:00...' -> 's1:1')

def on_packet(pkt, inport: 'external_ingress'):
  global_mac_table.insert(pkt.eth.src, inport, 500) # key, value, timeout

  if pkt.eth.dst in global_mac_table:
    return shortestPath(inport, global_mac_table[pkt.eth.dst]), pkt
  else:
    return spanningTree(inport), pkt

System requirements

You need a Linux OS to run the compiler and demo apps, typically Ubuntu Server 16.04 is preferred
Make sure you have python3 and python3-pip installed

Steps

Step1: Prepare for the mininet(BMv2) environment

please install bmv2 and mininet, and follow the p4 tutorial to get some helper script.

Step2: Clone the code to a directory

# git clone --depth=1 https://github.com/magellan-public/magellan.git

Step3: Install the dependencies and setup

# pip install -r requirements.txt
# python3 setup.py

Step4: Start mininet(BMv2) environment and ping from h1 to h2

start a mininet with 4 switches and 2 hosts, and ping from h1 to h2, at this moment, the ping command will fail.

# python run_exercise.py -t test/topology/l2-mn.json -b simple_switch_grpc
# h1 ping h2

Step5: Compile the L2 magellan application

compile the l2 test application to a directory, this step will generate initial p4 pipeline source code and runtime flow rules.

# magc -a test/apps/l2/on_packet.mag -t test/topology/l2-p4-4sw.json -o build

Step6: Run the program

start the l2 program as follows, then the ping command will success.

# mag_run -d build

guodong/magellan-pcl