RE: Newbie Question about addressing impacts

["Fleischman, Eric" <eric.fleischman@boeing.com>]

This is an interesting idea. However, if we embed proxy functions into border routers it would potentially add overhead (as well as latency) and make them harder to manage. Specifically, the number of border routers is likely to increase as network perimeters become more porous. Thus, this idea carries with it the need to ensure that these distributed routers can be configured with consistent policies.

Simple is good in operations.

-----Original Message-----
From: marcelo bagnulo braun [mailto:marcelo@it.uc3m.es]
Sent: Thursday, August 12, 2004 1:59 AM
To: Iljitsch van Beijnum
Cc: Multi6; Fleischman, Eric
Subject: Re: Newbie Question about addressing impacts



El 11/08/2004, a las 19:36, Iljitsch van Beijnum escribió:

> Hi Eric,
>
> On 11-aug-04, at 18:13, Fleischman, Eric wrote:
>
>> Newbie question: Assuming a multihomed environment with N ISPs 
>> supporting a corporate network of 2K+ routers and 200K+ computers. Is 
>> multi6 proposing that the interfaces of the routers and computers 
>> each have N different global addresses, one set for each of the N 
>> ISPs?
>
> At this time, multi6 isn't exactly proposing anything. But yes, pretty 
> much everything that's under discussion right now assumes this.
>
>> If so, has multi6 considered the impact of this approach to IGP and 
>> EGP performance? If so, what was the consensus conclusion to that 
>> evaluation?
>
> There hasn't been an evaluation.
>
> There is a feature that I'm very much in favor of that could help 
> here: the ability to implement a multihoming solution in middleboxes 
> or border routers. In such a scenario, there would be a set of 
> addresses for internal use in the site, and the proxy multihoming 
> devices add the multihoming capability somewhere close to the border 
> of the network. The good thing here is that there is no need to modify 
> each individual host to obtain multihoming benefits, and it's easier 
> to implement policy in a few central places rather than distributed 
> over all the hosts in the network.
>


Well, even if you do a per host locator selection you can still manage 
policy in a centralized fashion, if we define proper mechanisms to 
perform policy distribution, for instance a dhcp option for 
distributing RFC3484 policy table or something similar. However, i 
don't know how you can enforce policy in a centralized manner when 
locator selection is performed by the end hosts.

anyway, i agree that proxys are attractive, especially for supporting 
advanced features like session survivability in legacy hosts

regards, marcelo
>