[ Date Index ][
Thread Index ]
[ <= Previous by date /
thread ]
[ Next by date /
thread => ]
Jon Lawrence wrote:
On Tuesday 13 January 2004 8:17 am, Neil Williams wrote:On Monday 12 Jan 2004 10:37 pm, Jon Lawrence wrote:I'm on eclipse and I see what I think is an odd route to 82.145.32.65 it's considerably different to the route to 82.145.32.2 even though it must go through the .2 address to get there.
Looks cool from my Demon SDU.
Yes. Isn't this a feature of TCP/IP redundancy? You never can tell which route packets will take?
True - worse the way back can be different from the way forward - so traceroute presents at best a simplified picture of reality. But Internet routing is pretty stable in the sense if Demon are routing transatlantic traffic via one carrier today they'll probably be using the same carrier in 10 seconds time. It is generally a very bad sign if packets frequently switch the route they are taking.
A way of avoiding bottlenecks instead of getting stuck. There are usually many available routes between any two points.
Evil rumour - whilst IP protocol might support that the Internet structure is brittle and with lots of branches being broken off by small problems (the bigger stuff run by the big carriers does have more redundancy but you have to get there first). There was a great article on analysing the structure of Internet routing in Scientific American a year or two ago. The issue is money - bandwidth is expensive - so letting someone else use yours because his is bust disadvantages your users - so it only happens by agreement or by paying. IP was designed for military networks where ultimately it was all under one (or a few) administrative command, and the vast majority of the nodes are reasonably expected to be friendly and co-operative. So nothing like it's current status at all.
It is a feature of BGP routing. However, my packets always seem to go via the same routes and those routes don't make sense. Neil, the routes taken from your machine are almost the opposite to mine. Looks to me like eclipse have 3 or more routes with the same metrics defined - arghh. Going via the US really screws up my voip.
Do we have anyone who does BGP routing related stuff fulltime? - my experience is Internet routing people often live their own little lives - just doing routing. The nice folks at NANOG can probably explain what you are seeing (although technically off topic), whether they can explain what to do about it to Eclipse is another question. [root@xxxxxx /root]# traceroute 82.145.32.65 traceroute to 82.145.32.65 (82.145.32.65), 30 hops max, 40 byte packets 1 anchor-du-17.access.demon.net (195.173.57.17) 65.878 ms 89.521 ms 78.400 ms 2 anchor-core-21-fxp3.router.demon.net (195.173.57.251) 69.277 ms 70.498 ms 68.948 ms 3 anchor-border-1-4-0-2-3.router.demon.net (158.152.0.185) 67.837 ms 88.050 ms 79.228 ms 4 linx1.teleglobe.net (195.66.224.51) 77.511 ms 89.757 ms 81.288 ms 5 if-0-0-0.bb2.London.Teleglobe.net (195.219.96.81) 75.686 ms 89.429 ms 78.876 ms 6 ix-3-1-0-822.bb2.London.Teleglobe.net (195.219.2.34) 98.120 ms 109.532 ms 98.939 ms 7 wi2.westloc.com (82.145.32.2) 97.934 ms 109.961 ms 98.754 ms 8 wc3-2.westloc.com (82.145.32.65) 98.339 ms 99.502 ms 108.331 ms 32.2 is one hop shorter - nothing odd from here.
Attachment:
pgp00080.pgp
Description: PGP signature