laputarobot/sdn-cp-loadbalancer

ElastiCon, an elastic distributed controller architecture is proposed in which the controller pool is dynamically grown or shrunk according to traffic conditions to address the load imbalance situations in the network.

Elasticon:

Ref - ElastiCon: An Elastic Distributed SDN Controller, Advait Dixit, Fang Hao, Sarit Mukherjee, T. V. Lakshman, Ramana Rao Kompella.

Present day networking has revolutionised itself by providing a centralized view of the entire network, making it easier to manage and control. Software Defined Networking is an emerging paradigm for implementing centralized control in modern networking technologies like data centers where the data is handled by a large amount of switches. Handling this large data dynamically is a challenge in modern day networking since performance is a vital criteria for every client. The key limitation of present day network architecture is the static configuration between the switch and controller which results in an uneven distribution of the load in case of sudden dynamic shift in the load . To address this problem, ElastiCon, an elastic distributed controller architecture is proposed in which the controller pool is dynamically grown or shrunk according to traffic conditions to address the load imbalance situations in the network. Depending upon the load, Elasticon automatically balances the load among the controllers thus ensuring normal performance even during peak traffic. We implement a database to determine the load on the controllers and also a switch migration protocol to transfer the switch to another controller for load balancing.

Code:

controller_1.py and controller_2.py: RYU Controller app with Elasticon Protocol. load_balancer.py: Load Balancer App for DB access and migration decision making. multi_ctlr_topo.py: Topology file for Mininet.

Steps to run:

1. Start mongodb using: sudo mongod
2. Launch Controller 1 using: ryu-manager --verbose --ofp-listen-host <IP address of controller 1> controller_1.py
3. Launch Controller 2 using: ryu-manager --verbose --ofp-listen-host <IP address of controller 2> controller_2.py
4. Update IP addresses of the two remote controllers at net.addController() in mininet topology file.
5. Launch mininet for remote controllers with the topology file: sudo mn --custom multi_ctlr_topo.py --controller=remote --topo mytopo

Steps to verify:

1. Open a python shell and connect to the db using

   client = MongoClient('localhost', 27017)
   db = client.elastiCon
   controllers = db.controllers
   flags = db.flags
   gen_id = db.gen_id
   cmf = db.cmf

2. Monitor the Packet counts at both controllers using controllers.find_one({'id':'1'})
   controllers.find_one({'id':'2'})

3. Initiate a ping in the mininet

   mininet> h1 ping h2 -i 3

   Here we use -i 3 so that packet counts can be clearly monitored.

4. Observe that, when the count reaches 150, the switch is migrated and the controller ceases to receive further packetINs. Instead, the other controller's packetIN resets and starts to increment.

5. Between multiple runs ensure to drop the documents of the db using:

   controllers.drop()
   flags.drop()
   gen_id.drop()
   cmf.drop()