Interesting Result:  When the packet loss rate is set to 1/6 ( ie, drop one out of every 6 packets ). The elapse time varies a lot.

tcpdump result shows that there's some time gap between sending packets,  ie. the sending host "madrid" slows down.

Iteration Times Time elapse (Second)
1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

Random

1st 2nd
10000 1.77 12.73 1.77 1.77 1.78 1.77 1.77 346.65 1.78 1.77 122.34 114.58
20000 3.35 3.35 3.35 3.35 3.34 3.34 3.34 3.34 3.34 455.36 181.67 189.52
30000 4.91 4.92 4.92 4.92 1971.68 4.92 652.67 ........ 313.62 328.32
40000 6.45 6.45 6.45 6.44 6.45 1129.76 ........ 385.51 392.27
50000 8.01 1597.12 8.01 1008.76 566.48 8.01 ........ 611.35 528.09
60000 9.58 9.57 9.57 2006.39 2993.67 9.60 ........ 583.14 614.00
70000 10.95 4946.53 1128.08 4944.91 ........ 779.36 824.49
80000 12.71 12.71 4000.19 211.84 ........ 765.21 819.63

"........":   I had planed to do 10 times experiments for each "Iteration Times". but it take too long time at "......" So I gave up, and start new experiments.

The result shows that the first experiment always take less elapse time

The tcpdump(tcpdump dst host 192.168.2.2 ) result on router (dublin) is: dumpRegular1.gz

The above data/code are on london.cs.unh.edu:~lin/Dec4. Since london is behind the firewall, you can browse it from its mirror site:

student3-758/linuxproject/Dec4.

Madrid Raw experiments Elapse Time Data,   Regular    Random

Dublin (router) packet Loss records. Regular     Random

Regular Packet Dropper code

Random Packet Dropper code