1. 05 Jul, 2007 4 commits
  2. 02 Jul, 2007 4 commits
  3. 28 Jun, 2007 3 commits
  4. 27 Jun, 2007 4 commits
  5. 26 Jun, 2007 4 commits
    • Sachin Goyal's avatar
      · 4f6ed4f3
      Sachin Goyal authored
      Setting the sql mode for the database session to 'STRCIT_ALL'.
      This is to handle the database setting difference between mysql on Ops
      and mysql on DataPository.
    • Sachin Goyal's avatar
      · 5ebb54de
      Sachin Goyal authored
      Added an extra layer of protection to prevent to-and-fro switching
      between different nodes on a site due to CPU USAGE difference.
      Previously if a node has less cpu load than current best node (discounted
      by a threshold of 10). Now this has to happen 2 times consecutively
      for the switching to happen.
    • Sachin Goyal's avatar
      · 9b756cf8
      Sachin Goyal authored
      Added randomness for initial starting of bw probes scheduled by
    • Pramod R Sanaga's avatar
      Changed the cross-correlation to use FFT - makes it really fast! · f80056e0
      Pramod R Sanaga authored
      Cross-correlation calculation now takes less than 15 sec for a 15 minute
      TCP transfer - averaging time is still huge though...
      Added documentation to explain what is going on and how a run
      is declared correct/incorrect.
  6. 25 Jun, 2007 2 commits
  7. 22 Jun, 2007 2 commits
  8. 21 Jun, 2007 3 commits
    • Pramod R Sanaga's avatar
      · 2a8299ee
      Pramod R Sanaga authored
      SanityCheck.py is the script for checking cross-correlation of Flexlab
      runs. It works for both TCP and UDP transfers. I also changed
      auto-pelab.ns and added a "-t" option to start-experiment
      and stop-experiment to enable these sanity checks.
      Note: I haven't figured out how to kill sudo tcpdump sessions
      on PlanetLab with program agents - so, for now they have to be
      killed manually after a stop-experiment. Will change this after
      talking to Mike.
      The cross-correlation code works correctly - I may need to tweak it
      a bit in the near future for averaging times.
      Also, it puts too much load on Ops and (i think) takes too much time.
      I will change it so that all the correlation calculations are done on
      the experimental nodes.
    • Robert Ricci's avatar
      Give a more meaningful error message if the user tries to upload a · 9812a826
      Robert Ricci authored
      file that does not exist.
    • Russ Fish's avatar
      Mention tested on 6.2, fix minor typos. · 832661c4
      Russ Fish authored
  9. 20 Jun, 2007 9 commits
  10. 19 Jun, 2007 5 commits