dheeraj@cs.umd.edu (Dheeraj Sanghi) (09/13/90)
Recently, we have seen massive reordering of packets in the internet. Previously, when we noticed reordering, it was always 2 packets who have been sent together, reach in the opposite order. But now we are seeing as much as 10 packets sent at the rate of 1 packet per clock tick (10 ms) reaching exactly in the opposite order. We have not yet figured how to reproduce it at will, but we have seen it quite often. Here is some data from our observations. PKT# Sendtime Recvtime Delay 3 160 370 210 4 170 360 190 5 180 360 180 6 190 350 160 7 200 350 150 8 210 340 130 9 220 340 120 10 230 330 100 11 240 320 80 12 250 320 70 13 260 400 140 (All times in ms.) We are using DTP, an in-house protocol with the same functionality as TCP. We are using IP, so it shouldn't really matter what transport layer protocol we use. The route being taken is (from the traceroute program): rhodes.cs.umd.edu 128.8.128.1 mrb-gw.umd.edu 128.8.109.1 bogon-gw.umd.edu 128.8.253.7 128.167.102.2 128.167.104.1 sura1.sura.net 128.167.1.1 We are using IP_LSRR option to send packet to sura1.sura.net and back. Both sender and receiver programs are on huginn.cs.umd.edu (128.8.128.35). The acknowledgements are sent within machine. We have also tried sending packets upto various intermediate gateways and have seen reordering from all gateways except rhodes and mrb-gw. (The first two gateways.) Any ideas as to what is causing the whole block of packets to reorder and reach the destination in exactly the opposite order. Dheeraj Sanghi (dheeraj@cs.umd.edu) Ashok Agrawala (agrawala@cs.umd.edu) -- Dheeraj Sanghi (h):301-794-6247 (o):301-405-2723 Internet: dheeraj@cs.umd.edu UUCP: uunet!mimsy!dheeraj When you find you are on the side of the majority, it is time to reform. - Mark Twain