- Before every exercise, read the appropriate part from "Chapter 2". - Your conclusions reveal what you understand from the traffic (based on the statistics that the exercise has told you to determine). Do not merely work towards doing just the plots and statistics that the exercise points out! Work towards your conclusions and understanding of the traffic! This is important. - If a statistical property has parameters, experiment with different parameter values. If "Chapter 2" gives sugggestions on how to determine the parameters, try these also. - Work together. Organize and delegate within your group. This will save you time. Remember to acknowledge contributions! - Exercise 2: Parameters to be tweakeed are found at least for the following * Window size in Arithmetic Moving Average * Smoothing constant for Exponential Moving Average * Bin widths for distributions Refer to "Chapter 2". - Be specific in your report on the definitions of AMA, EMA (or is it EWMA) etc. - To be sure what you are doing, it is suggested (and strongly recommended) to implement AMA, EMA, EWMA etc. instead of using the methods provided by Matlab. - Some of you have started to wonder why the flow arrival process in funet-traces is as it is. You can start to search the truth based on the fact that the flow entry in the funet-trace file appears only _after_ the flow has finished (seen its last packet). There is no (or very little) correlation to the start time (between start time and end time, that is). Should you want to understand more on the phenomena, please filter the flow-start-packets (marked with '0' in the trace) and their timestamps from dec-traces and observe the flow arrival process there (and compare to the arrival process in funet-traces). - I have also prepared a large (3.2 Gbytes) file containing all timestamps of the flow arrivals and departures (not sorted) of the funet-trace from 24 hour period. If you wish (and are ready to do some analysis) you can get access to this file. Contact Mika for more details. Due to the large size, the Matlab-sampling is probably out of the question and the initial sampling should be done with C or Perl etc. - Exercise 3: Implement cross-correlation function in Matlab according to the definition found in "Chapter 2". Matlab's xcorr lacks in functionality.