[c-nsp] Traffic Engineering Internet links automatically

[Oliver Boehmer (oboehmer)]

BGP multi-path (e+iBGP) along with dmzlink-bw would do the trick to
perform unequal cost load-sharing over those links, don't know if
Juniper supports this or a similar feature for your upstream ISP to do
the same on their end. 
In which direction are your links filling up? Inbound, I guess?

	oli

Kim Onnel <> wrote on Monday, March 27, 2006 3:09 PM:

> I am sure someone has been in my place before and there is something
> that can be done on this matter, i'll try to explain again in more
> details, maybe i missed something, i hope i can get suggestions on
> Best practices for this 
> 
> 
> upstream \
> upstream ---- multiple links ----- GW-INTERNET---
> upstream /
> 
> the multiple links are all terminated on our side at 1 router
> (GW-INTERNET) and the upstream isnt 1 router, but same AS
> 
> Lets say i have 3 PoS and 2 Giga
> 
> My manager wakes up each day, checks MRTG, finds that the Giga has
> some room and that PoS is full, so he tells me to move some networks
> around to achieve this, as Gert said above, it is time consuming,
> specially when moving 1 /24 doesnt do it and half of my day is spent
> in excel sheets and updating everyone ?
> 
> What is the correct thing to be done here ?
> 
> On 3/27/06, Gert Doering <gert at greenie.muc.de> wrote:
>> 
>> Hi,
>> 
>> On Mon, Mar 27, 2006 at 01:33:25PM +0200, Kim Onnel wrote:
>>>> On Mon, Mar 27, 2006 at 12:11:39PM +0200, Sami Joseph wrote:
>>>>> What does ISPs do in this situation, do you guys just manually
>>>>> keep moving/advertising networks between diff. peers ?
>>>> We get bandwidth in sufficiently large chunks...
>>> 
>>> You mean that you have enough that you dont need to keep moving ?
>> 
>> Yes.  Dedicating vast amounts of engineer time to work on "perfect"
>> load balancing can easily get much more expensive than getting a
>> bigger pipe. 
>> 
>> (It all depends on circumstances, of course.  In Germany, uplink
>> bandwidth is fairly cheap these days, and if you're located in the
>> right places, it's just a matter of throwing a fiber link across the
>> building, and getting a GigE uplink port with some 100 Mbit/s. of
>> "committed" bandwidth) 
>> 
>>> (And in certain cases we move around traffic to/from certain remote
>>> ASes 
>>>> by prepending inbound and using communities/prepends outbound)
>>> 
>>> Well, in our case, we want to tailor it to do the maximum
>>> utilization possible, so any few megabits is useful, i dont think
>>> that prepending and communities are the way ? 
>>> 
>>> Any known methods ?
>>> Is OER used for this ?
>> 
>> OER will do for outgoing traffic.
>> 
>> For incoming traffic, you'll have to live with what BGP can give you
>> - prepending, communities (to have your upstream prepend to certain
>> peer ASes of this upstream), etc. 
>> 
>> gert
>> --
>> USENET is *not* the non-clickable part of WWW!
>> 
>> //www.muc.de/~gert/
>> Gert Doering - Munich, Germany
>> gert at greenie.muc.de
>> fax: +49-89-35655025
>> gert at net.informatik.tu-muenchen.de
>> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/