**This is an old revision of the document!** ----
====== Scenario Proposal ====== The main metric of interest is total end-end transmission duration; our goal is to minimize this metric. A secondary metric is fairness: a solution that has low transmission duration but saturates the network (causes packet loss on existing flows) may not be acceptable. The following are parameters which we do not expect to vary much, as they are dependent on the network configuration that our solution is tailored for (also, it is difficult for us to change these parameters): - RTT (50,75,100,125 ms) - packet loss (??) - throughput (1Gb/s, 10Gb/s) - number of hops (10,15,20,25) - intermediate queue sizes (??) - end host disk speeds (??) The following are factors which we will vary across our experiment in order to find an ideal solution. We will do a full factorial experiment. - solution application (GridFTP, paraFetch, FDT) - number of parallel streams (i.e. 4 8 16 20) - host TCP settings: - queue sizes (64k - 100MB) - TCP backoff: bic, cubic, vegas - with/without encryption - memory-memory/disk-disk
