roy@phri.UUCP (Roy Smith) (08/16/89)
Can somebody tell me how this happens: [...] 64 bytes from 132.247.5.1: icmp_seq=14. time=2418. ms 64 bytes from 132.247.5.1: icmp_seq=15. time=2570. ms 64 bytes from 132.247.5.1: icmp_seq=15. time=2613. ms 64 bytes from 132.247.5.1: icmp_seq=16. time=1700. ms [...] ----132.247.5.1 PING Statistics---- 51 packets transmitted, 49 packets received, 3% packet loss round-trip (ms) min/avg/max = 1612/2111/3479 Note that ping 15 got answered twice. What causes that? It doesn't seem like a fluke -- I tried pinging 132.247.5.1 three times and each time got a doublet within the first 10 or 20 packets. The machine in question is in Cuernavaca, Mexico, with (to the best of my knowledge) a terrestrial microwave link to Mexico City and some satellite link from there. You should see what telnets look like. -- Roy Smith, Public Health Research Institute 455 First Avenue, New York, NY 10016 {att,philabs,cmcl2,rutgers,hombre}!phri!roy -or- roy@alanine.phri.nyu.edu "The connector is the network"
Mills@UDEL.EDU (08/17/89)
Roy, You may have caught it in an ugly mood. Using ICMP Timestamp messages, it has a most satisfying and stable delay of 1609-1616 ms delay and clock jitter of 8 ms max. With more data as yours, I would look for clusters of delay separated by 270 ms or so, the roundtrip satellite delay. Dave
karn@ka9q.bellcore.com (Phil Karn) (08/19/89)
In article <3936@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: > > Note that ping 15 got answered twice. What causes that? It >doesn't seem like a fluke -- I tried pinging 132.247.5.1 three times and >each time got a doublet within the first 10 or 20 packets. [...] A few days ago I was pinging my way around the NSFNet/Internet, measuring round trip delays and comparing them to some expected values I had computed. I too noticed that quite a lot of packet duplication was going on. When pinging berkeley.edu from rotgut.bellcore.com, I consistently saw a 10% packet duplication rate. Traceroute showed a perfectly reasonable path from bellcore-net through JvNCNet, NSFNet and BARRNET to Berkeley. This got me curious, so I started to isolate the problem. I could ping JvNC's last router (the one adjacent to their NSF backbone switch) with no packet duplication and virtually no loss. But when I pinged the NSS at Merit, just one hop away from JvNC, I began to see duplicate packets. I therefore concluded that the duplication was happening inside NSFNet. Does anyone have any independent knowledge of this, or can anyone offer an explanation of what mechanism could be responsible? It's hard to understand how packets can get duplicated in the Internet unless there is a retransmitting link-layer protocol somewhere. Phil