TELCOM 2120 Network Performance
Homework 2 Spring 04
Problem 1
A sorted list of the measured call set up delay for a multicast switched VC connection in a ATM network is shown
below with the data in msec. The experiment was repeated 45 times assuming the data forms a random sample
802 809 812 814 817 819 823 855 862 886 883 892 896 896 900 904
905 930 932 933 948 958 959 963 964 968 970 970 972 976 981 981
985 989 990 991 1008 1012 1022 1028 1033 1034 1046 1048 1050
(a) Determine a 90% confidence interval on the mean call set up delay.
(b) What is the relative precision of the confidence interval
(c) Determine a 90% confidence interval on the .5 quantile value of the call set up delay
Problem 2
From the Jain Textbook Problem 13.2 parts (b), (c), and (d)
Problem 3
Collect ping delay data from Pitt to www.humboldt.edu. This can be accomplished by pinging
www.humboldt.edu using a 64 bit data packet. Repeat the ping at least 500 times (note may end up with less
than 500 measurements due to packets being dropped).
(a) Check the independence of the data by constructing either a scatter diagram for the data one apart, then
repeating for two a part etc. or plotting a correlation plot. Turn in your data with the plots. If you are
collecting the data from a unix or solaris machine the following may help you get the delay data into a file
for analysis. Put the following lines into a file called 'myping'.
#!/bin/csh -f
if (($1 == "") || ($2 == "") || ($3 == "")) then
echo "Usage: myping <machine name> <pkt size> <count>"
endif
echo "Trying.... ping -s $1 $2 $3"
ping -s $1 $2 $3 | grep time | awk -F= '{print $3}' |awk -F. '{print $1}'
Then make the file an executable (chmod u+x myping).
to run it: "myping <machine> <pkt> <count> > output_file"
Example of running it
myping hiMolde.no > pingdata
(b) Determine some summary statistics of the data (mean, median, variance, range)
(c) Construct a boxplot of the data and comment on the results
(d) Plot a histogram of the data and state which theoretical distribution do you think might fit the data (you
don't need to try and fit the data just state a distribution).
