* Why doesn't BGP pick the link with the highest bandwidth, or, better * yet, pick the link with the highest bandwidth AND least congestion to * label as the "best" available route? The needed information is avail-
The first one is easy, in fact you can do that yourself by fiddling with metrics or such on the different BGP sessions. The second one would have dramatic consequences in terms of route instability. You pick one route now because of load on the link, the load changes and you pick the other, now BGP will have to change the announcement of this network to other peers. So, now we not only have flaps because of links/routers going up and down, we also have flap because of load changes on the network. The result: you are dampened out forever, or the network falls over.
-Marten
On the radio1:: "101 north is congested, take 280 instead" Guess what hapens 5 minutes laster; On the radio2:: "280 north is congested, take 101 instead" goto on the radio1 --Peter
"101 north is congested, take 280 instead" Guess what hapens 5 minutes laster; "280 north is congested, take 101 instead"
Sort of like the cruise control in my car. The car starts going too fast and it *eases* off. The car starts going too slow and it adds a *little* more throttle. My cruise control works quite well. So could routers.
"101 north is congested, take 280 instead" Guess what hapens 5 minutes laster; "280 north is congested, take 101 instead"
Sort of like the cruise control in my car. The car starts going too fast and it *eases* off. The car starts going too slow and it adds a *little* more throttle.
My cruise control works quite well. So could routers.
Your cruise control doesn't have to communicate with 80 other cruise controls, each of which communicates with 80 other cruise controls, each of which then has to decide whether they have to change any course or acceleration settings based on the updated info. :) Avi
On Mon, 11 Nov 1996, Avi Freedman wrote:
Sort of like the cruise control in my car. The car starts going too fast and it *eases* off. The car starts going too slow and it adds a *little* more throttle.
My cruise control works quite well. So could routers.
Your cruise control doesn't have to communicate with 80 other cruise controls, each of which communicates with 80 other cruise controls, each of which then has to decide whether they have to change any course or acceleration settings based on the updated info.
CSMA/CD :-) Michael Dillon - ISP & Internet Consulting Memra Software Inc. - Fax: +1-604-546-3049 http://www.memra.com - E-mail: michael@memra.com
Sorry to add more noise... but here's a thought: On Mon, 11 Nov 1996, Avi Freedman wrote:
Your cruise control doesn't have to communicate with 80 other cruise controls, each of which communicates with 80 other cruise controls, each of which then has to decide whether they have to change any course or acceleration settings based on the updated info.
I'm basically a stub network. I take routes from 3 AS's right now, but I don't redistribute anything I learn. The only redistribution that goes on for me is in my iBGP peering. The convincing argument for me against doing the "intelligent" route selection has been related to the huge route flapping that would ensue. My question is, if I'm not redistributing the routes (except internally, and I have a certain tolerance for route flapping internally) then why not do the intelligent route selection? (the route selection would of course have dampening parameters etc.. and be turned off by default, but have the ability to be enabled, etc) Ed Henigin ed@texas.net
I'm basically a stub network. I take routes from 3 AS's right internally) then why not do the intelligent route selection?
Because some predict that it will cause the death of the Internet? You should do it (in some manner slow enough to not cause problems on your internal net).
My cruise control works quite well. So could routers.
IGRP/EIGRP has a knob that you can turn on to have it take congestion into account when making routing decisions. Its off by default. You know why its off? Things get really unstable otherwise. But if someone really really wants to play with this sort of thing, go ahead & turn it on - on your *own* network. Make it work there first - a couple of Phd dissertations later & folks may be interested in turning it on in their own networks & maybe in the Internet at large. --asp@partan.com (Andrew Partan)
participants (6)
-
Andrew Partan -
Avi Freedman -
Edward Henigin -
jon＠branch.net -
Michael Dillon -
Peter Lothberg