We probe CAIDA monitor champagne.caida.org (located on UIUC network) with traceroute (UDP) packets of increasing size. The path which connects CAIDA network with the monitor goes through vBNS, which at the moment of writing has OC-12 and OC-48.
Each experiment sends 10 traceroute packets to each of the 10 hops connecting CAIDA to the UIUC monitor.
The first experiment tests coarse-grained banwidth-dependent network properties. A total of 65 traceroutes with packet data size of 1,2,...65 kilobytes is sent. The RTTs of these packets are shown on the graphs below. Same experiment was run 6 times, using hosts crunch, lib5, chianti, cider and ipn (twice).
A typical traceroute (as run from hosr crunch) looks as follows:
/usr/sbin/traceroute -q 10 champagne 1000 1 pinot-fe2-0-0 (192.172.226.65) 1.041 ms 1.018 ms 0.889 ms 0.942 ms 0.860 ms 0.902 ms 0.901 ms 0.901 ms 0.903 ms 0.912 ms 2 cs-f-vbns.sdsc.edu (198.17.46.43) 5.769 ms 5.215 ms 5.447 ms 5.488 ms 5.459 ms 5.353 ms 5.324 ms 5.491 ms 5.367 ms 5.445 ms 3 jn1-at1-0-0-2.rto.vbns.net (204.147.129.126) 5.367 ms 5.655 ms 5.323 ms 5.482 ms 5.373 ms 5.445 ms 5.322 ms 5.487 ms 5.658 ms 5.812 ms 4 jn1-so7-0-0-0.hay.vbns.net (204.147.128.81) 17.006 ms 17.205 ms 17.280 ms 16.833 ms 17.247 ms 16.790 ms 17.450 ms 17.287 ms 16.825 ms 16.920 ms 5 jn1-so7-0-0-0.dnj.vbns.net (204.147.136.141) 40.366 ms 40.462 ms 40.388 ms 40.390 ms 40.421 ms 40.383 ms 40.511 ms 39.936 ms 40.015 ms 40.221 ms 6 jn1-so7-0-0-2.dng.vbns.net (204.147.136.144) 64.673 ms 64.843 ms 64.462 ms 64.428 ms 64.468 ms 64.471 ms 64.349 ms 64.426 ms 64.842 ms 64.423 ms 7 cs-atm0-0-0-25.ncsa.vbns.net (204.147.136.90) 67.658 ms 68.045 ms 67.745 ms 68.239 ms 67.908 ms 67.993 ms 67.827 ms 67.766 ms 67.931 ms 68.088 ms 8 charlie-atm1-0-5.ncsa.uiuc.edu (141.142.11.2) 68.602 ms 68.753 ms 68.404 ms 68.239 ms 68.564 ms 68.442 ms 68.406 ms 68.236 ms 68.688 ms 68.565 ms 9 141.142.0.2 (141.142.0.2) 68.445 ms 68.808 ms 68.729 ms 69.714 ms 68.567 ms 68.770 ms 68.525 ms 68.398 ms 68.896 ms 68.727 ms 10 champagne (141.142.121.4) 68.451 ms 68.506 ms 68.719 ms 68.565 ms 68.443 ms 68.402 ms 68.404 ms 68.524 ms 68.281 ms 68.404 msThe cities and areas passed by the path can be inferred from host names. This inference is corroborated by the vBNS maps of network topology and bandwidth map of Aug/21/2000 and prediction for Dec/31/2000. It has to pass metropolitan areas of Los Angeles, San Francisco - Bay area, Denver Colorado and Chicago on its way to NCSA and UIUC. The nominal bandwidth of the links along the path is OC-12 (622 Mbps), which may be of ATM or packet-over-SONET (POS) variety, or OC-48 (2.4 Gbps) POS.
According to RFC 1812, a router should send reply "TTL expired" with the address of the interface on which the packet was received. According to the same RFC, it should put the address of the interface on which the packet is sent when updating Record Route option in the IP header. This makes it possible to identify interfaces which belong to the same box, by sending traceroute packets to IP addresses and checking whether the source address in the response is different from the original destination. This idea was successfully used in the iffinder program developed by Ken Keys at CAIDA. Yet another way to identify interfaces is by using IP Record Route option and comparing recorded route to the interfaces returned by the traceroute. Recorded route is of course limited to 9 packets, nonetheless it is useful in many cases since the number of hops is either less than 9, or some of the intermediate hops do not insert their addresses in the option field.
Bandwidth shift which attains min.spread between minimum and median of resulting dtr. (step=1 ns/bit = 8 ns/byte) hop min.spr.at diff,ns/bit bw.est,Mbps 1 39.000 39.000 25.641 2 67.000 28.000 35.714 3 87.000 20.000 50.000 4 85.000 -2.000 0.000 5 90.000 5.000 200.000 6 93.000 3.000 333.333 7 102.000 9.000 111.111 8 124.000 22.000 45.455 9 121.000 -3.000 0.000 10 126.000 5.000 200.000 Bandwidth shift which attains min.spread between minimum and median of resulting dtr. (step=0.25 ns/bit = 2 ns/byte) hop min.spr.at diff,ns/bit bw.est,Mbps 1 38.750 38.750 25.806 2 66.500 27.750 36.036 3 87.000 20.500 48.780 4 85.000 -2.000 0.000 5 89.750 4.750 210.526 6 93.500 3.750 266.667 7 101.500 8.000 125.000 8 124.750 23.250 43.011 9 120.750 -4.000 0.000 10 126.250 5.500 181.818 Bandwidth shift which attains min.spread between minimum and median of resulting dtr. (step=0.0625 ns/bit = 0.5 ns/byte) hop min.spr.at diff,ns/bit bw.est,Mbps 1 38.688 38.688 25.848 2 66.438 27.750 36.036 3 87.062 20.625 48.485 4 85.000 -2.062 0.000 5 89.625 4.625 216.216 6 93.438 3.812 262.295 7 101.438 8.000 125.000 8 124.625 23.188 43.127 9 121.562 -3.062 0.000 10 126.250 4.688 213.333
Bandwidth shift which attains min.spread between minimum and median of resulting dtr. (step=0.0625 ns/bit = 0.5 ns/byte) hop min.spr.at diff,ns/bit bw.est,Mbps 1 24.125 24.125 41.451 2 50.500 26.375 37.915 3 51.625 1.125 888.889 4 67.500 15.875 62.992 5 36.812 -30.688 0.000 6 62.875 26.062 38.369 7 56.375 -6.500 0.000 8 29.250 -27.125 0.000 9 81.375 52.125 19.185 10 94.500 13.125 76.190Data collecting and processing scripts.