/SDN-Network-Statistics-Collection

Collecting instantaeous statistics of the network (SDN) - using Mininet + ODL

Primary LanguagePython

Statistics collection to perceive the instantaneous behaviour of an SDN enabled network

Problem Definition

The state of the art congestion control algorithms are based on End to End packet statistics. Leveraging the network feedback for congestion control is another promising dimension, but the catch is the availability of network statistics at one place. Software Defined Networking can mitigate this limitation and hence be a pillar for network suggested TCP Congestion control.

How does it work? and Versions

A topology created in Mininet having some TCP flows happening between hosts in the topology. User queries a centralised server for getting the needed statstics. The statistics collection part has 2 versions as mentioned below.

  1. StatCollection from all switch interfaces using Python script.
  2. Using the statistics collected by ODL.

Steps

  1. As TCP connection takes place, the sniffer finds the path taken by the connection.
  2. In the background statistics are being collected for all interfaces (of all switches).
  3. When the user queries for a particular connection,
    1. The list of intermediate switches which were used by the connection is found out.
    2. The statistics for each of these interfaces are extracted and sent back to the client.

Modules

1. Topology Generation Source code

  • A generic topology generator.
  • Input is a CSV file whose 1st column specifies all nodes
    • Hosts' name should start with 'h'
    • Switches' name should start with 's'
    • Controllers' name should start with 'c'
  • Since the internet can have different controllers controlling different part of network, we have support for different controllers. With a many to one mapping between Switch and controllers.
  • Which controller controller controls which switch can also be specified.
  • 2nd column onwards specifies the nodes that are neighbours of the node in th e corresponding row's first column.
  • Currently the host-switch links have 5Mbps link with a delay of 5ms and the switch to switch links have 100Mbps link with 7.5ms delay.
  • Make sure you are in the corresponding directory + followed all Set-up instructions
  • Run the script using sudo python Generator.py
  • Once the script runs, it generates 4 different files (3 Python scripts and 1 CSV file).
    • topo_consolidated.py
    • stat_consolidated.py
    • sniffer_consolidated.py
    • Switch2PortMapping.csv

2. Sniffer file Source code

  • This file is generated by the previous step.
  • Sniffs TCP segments from all the active interfaces of all switches in the Mininet emulation environment.
  • Every packet is processed to check if the hextuple (SourceIP - Destination IP - Source TCP Port - Destination TCP Port - SwitchID - Switch's PortID) is unique.
  • If unique, it is recorded by sending to MySQL running at an external system.
  • This helps in figuring out the path taken by each TCP flow.

3. StatCollector - Version 1 Source code

  • Internally runs the command ovs-vsctl get Interface <interface_name> statistics
  • Parses the output and pushes to MySQL server.

4. Web service to query MyService_SDN SDN_frontend

  • Run the SDN_frontend and MyService_SDN on the tomcat server.
  • Change the url,username and password in ConnectionProvider.java inside dBConn of MyService_SDN to connect to the MySQL server.
  • Launch the Odl.html page of SDN_frontend.
  • Enter the source ip address and the destination IP address.
  • It displays a list of intermediate switches,which is the path used by a packet to travel from the given source to destination
  • On selecting a particular node,it displays the overall statistics which includes all the ports the node has.
  • It also displays the names and statistics of the ports which the packet travels through in the given node.
  • Further,the statistics at port level is also displayed.
  • Using these statistics collected at node level,the user at the source can modify the TCP congestion window dynamically.

5. StatCollector - Version 2

  • When the user enters the source and destination IP,we retrieve a list of intermediated nodes,it uses in its path.
  • Using a HTTP request we connect to the ODL using a servlet which has the required authentication(username,password of the ODL controller).
  • We retrieve all the required statistics from the ODL in the form of JSON,parse it and store the values to MySQL server.
  • These values are retrieved and displayed to the user,which is similar to the 'Web service to query' in Version 1,step 3 onwards.

Running the project

  • sudo python topo_consolidated.py

  • xterm <any switch name>

  • sudo python sniffer_consolidated.py

  • In case you want to use the version 1 of this project,

    • xterm <any switch name>
    • Make sure MySQL server is running and accepting connections
    • In the xterm window run sudo python stat_consolidated.py

  • In case you want to run the version 2 of this project

    • Make sure the ODL server is running.
    • Change the username and password in ODL.java inside SDN_frontend to connect to the ODL server
    • You need not run sudo python stat_consolidated.py as the stats are collected by a HTTP request when we enter the source and destination.

  • To make use of the rest service directly without form

    • Change the Formparam to QueryParam in the functionname of each service to send the required parameters in the Url.
    • Change the annotations from @POST to @GET.
    • Connect to the web service using the suitable Url and pass the parameters(eg:src_ip,dest_ip)
    • The output is in the form of JSON.

Project delegation

  1. Statistics collection and SDN-Node communication – Sabari Nathan A. (MT2016120). @sabarinathana
  2. Creating web interface between client and SDN controller for information retrieval – Maneesha S. (MT2016119). @maneeshashivakumar