Nanog list uncharacteristically quiet?
Chris A. Icide
chris at nap.net
Sat Dec 14 19:59:15 UTC 1996
At 11:24 AM 12/14/96 -0800, Chris Caputo wrote:
>Here's something to chew on. As an end-node site who has recently become
>triply-homed I've been wondering whether it would be useful if routers had
>a way of making a route selection based on output queue levels. Ie.,
>let's say I have 2 paths to a destination of equal as-path distance
>through neighbor A and neighbor B. Based on recent discussion it sounds
>like IOS will send the packet to the neighbor with the lowest IP address.
>This is nice and predictable, but I wonder if it would be helpful to have
>an option that makes this decision be based on current output load of the
>link to each neighbor rather than IP address.
>Note, I am not suggesting in any way that this alter advertisements and
>causing flapping, I am only wondering about outgoing packets.
The problem that jumps out immediately is that of oscillating loading of
an interface. As soon as one interface becomes more loaded than the
other, what decision process are you going to make to move only a small
amount of traffic to the other interface? If you don't do this in small
then you will surely experience an oscillation between the two (or three)
interfaces as the lowest load interface immediately becomes the highest
loaded interface due to the decision to send traffic that way instead. Also,
remember that not only do you have to make the decision on your side,
but you also must let your peers know that you want your incoming
traffic to move also. So you now must have some method of advertising
your desires to the rest of the world.
Currently your best choice is to manually develop route-maps which will
balance both incoming and outgoing traffic on each interface. I definately
have some thoughts on the above idea, and how something like that may
be accomplished, but they are so very undeveloped that this is not the
time or place to display my lack of knowledge. I'd prefer to put it all down
on "paper" somewhere and then submit the idea if anyone who really
understands all this is interested. I'm sure none of my ideas are original.
Chris A. Icide
More information about the NANOG