May, 25th - May, 31st

We have updated the Load Balancer to consider the use case where a job was being incorrectly burst to the public Cloud by adding a tolerance value to the upload period threshold. The rest of the week was dedicated to collecting results and testing our solution with real network traffic, as opposed to network traffic generated by Iperf.

During the tests with real traffic, we came to the conclusion that there was something wrong with the PcapInputFormat used by p3, the open source PCAP file analysis tool for Hadoop. Map tasks were performing very differently among each other. After a thorough debugging session, Professor Ricardo Morla ended up developing our own PcapInputFormat.

NEXT OBJECTIVES: Continue collecting the results with Iperf and start writing the Results chapter in the dissertation document. Test the new PcapInputFormat and report back to the Professor.