Testing the fcast application that uses ALC reliable multicast

All scripts are in gw.orbit-lab.org:/home/kishore/Test-MCL (~/ in the rest of the document is equivalent to /home/kishore).

Important scripts

• start-mcasttx.sh - This is run on the sending node.
• start-mcastrx.sh - This is run on all receiving nodes.
• test-512B.sh - This is the main script that runs the above scripts on the sending and receiving nodes as well as collects data on the experiment run.

Steps to run Experiment

• Image nodes on the grid - [3,1] and all nodes from [4,1] to [15,20] (except [5,10] which is a dead node). There is a ruby script in my home directory called imageSubset.rb (which images all nodes from [4,1] to [15,20] except [5,10]), which you can use.

imageNodes [3,1] baseline-fcast.ndz
ruby imageSubset.rb

• Turn these nodes ON -

wget -O - "http://cmc:5012/cmc/nodeSet?nodes=[4..15,1..20]"
wget -O - "http://cmc:5012/cmc/nodeSet?nodes=[3,1]"

• Once the nodes are ON,
  • cd ~/Test-MCL
  • mkdir directory-for-this-run -

    mkdir 240nodes-run12
    

  • Make sure the sender (node3-1) has the directory you wish to transfer using multicast (for e.g. ~/Test-MCL/512B contains 10000 512 byte files, ~/Test-MCL/512B-5000 contains 5000 512 byte files).

    cd ~/Test-MCL
    scp -r 512B-5000 root@node3-1:/root/
    

  • Modify start-mcasttx.sh according to the directory you want to transfer - for e.g. to transfer 512B-5000, start-mcasttx.sh should contain the following command:

    fcast -R -send -a225.1.2.3/2323 -v0 -ospeed -plan -force /root/512B-5000 &> $1 &
    

  • Copy start-mcasttx.sh to node3-1 (sender):

    cd ~/Test-MCL
    scp start-mcasttx.sh root@node3-1:/root/
    

  • Copy start-mcastrx.sh to all receiving nodes:

    cd ~/Test-MCL
    ./copyfile.pl start-mcastrx.sh
    

  • Start experiment:

    cd ~/Test-MCL
    ./test-512B.sh 240nodes-run12 &> 240nodes-run12/240nodes-run12.txt