wiki:jackSandbox

Version 1 (modified by peter.stickney@bbn.com, 7 years ago) (diff)

--

Project Overview all the code has been updated at https://github.com/umkcdcrg01/ryu_openflow

  1. OpenFlow: AAN/SDN integration with Hadoop
  2. Monitoring: How can you monitor the results of "1" (i.e. compute resources, throughput)
  3. Opt-In: L2TP
  4. Security?
  5. Journal publication?

NB. All OF work will be conducted with OF 1.3

Milestones:

  1. Deploy Experiment at One GENI Site
  • Create Development Environment

a) Create rspec with 4 ovs switches and 4 VMs as follows:

                     h2
                      | 
                  -- sw2 -- 
                 |         |
          h1 -- sw1       sw4
                 |         | 
                  -- sw3 --
                      |
                     h3 

NB. h4 (not shown) will host your Ryu OF controller. h1 will be the hadoop master: h2 and h3 will be the slaves.

b) Install Ryu on h4 c) Review documentation/demos of Hadoop instances in GENI d) Install Hadoop on h1, h2, and h3 e) Document steps a)-d)

  • Develop SDN Applications

a) Integrate topology discovery to the learning switch module packaged with Ryu (simple_switch_13.py) b) Add path discovery (i.e. shortest, longest etc) mechanism to topology discovery module. c) Integrate flow statistics module to track through put of source-destination flows d) Document steps a)-c)

  • Develop AAN Applications

a) Develop light-weight script to monitor CPU and storage resources on hadoop slave nodes b) Document a)

  • AAN/SDN Integration

a) Use Ryu's northbound API, to provide network/topology info to hadoop. b) Evaluate's hadoop's ability to intelligently trigger flows on a particular path

based on the network characteristics (including topology and BW) and node characteristics (i.e. hadoop slaves, including CPU and storage resources)

c) Document a)-b)

  • Integrate security mechanisms?
  1. Deploy experiment in GENI (i.e. at 4 GENI racks)