ietf
[Top] [All Lists]

RE: 2 interfaces on same subnet

2001-06-01 03:50:02
* Loadbalance feature is something usually implemented and controlled by
routing protocols using metric rules and not by the addressing scheem;
* On case of pure redundancy need (hs), in some systems is common a
definition of a virtual ip address shared by several interfaces - but under
this circunstances, the system should only answer to ARP requests only att
one interface at time for a specific ip address.


This are two common project requests that can be solved without breaking
simple logical and effective engineering rules.

Question:  why is it  logical rule?
-- case study:

Imagine the scenario -> one system with two interfaces : eth0: ip address
10.10.10.1/ 24 and the other 
eth1: 10.10.10.2/ 24.
We have 10 Ip packet to get out to the destination ip address 10.10.10.3.
What destination interface should be used? How to manage the percentage of
traffic that will get out at one or other interface? In case of interface
failure what is the reasonable decision to follow?

this are typical issues of routing protocols - the decision should be
taken by the routing protocol, and not by the addressing scheeme itself.


Regards,
Jorge Cardoso.
                -----Original Message-----
                From:   John Stracke [mailto:francis(_at_)ecal(_dot_)com]
                Sent:   Quinta-feira, 31 de Maio de 2001 17:22
                To:     ietf(_at_)ietf(_dot_)org
                Subject:        Re: 2 interfaces on same subnet

                Valdis(_dot_)Kletnieks(_at_)vt(_dot_)edu wrote:

                > On Thu, 31 May 2001 17:40:53 +0300,
EXT-Faycal(_dot_)Hadjiat(_at_)nokia(_dot_)com  said:
                > > I search why we can't have two ethernet interfaces(with
different IP address
                > > of course) on the same subnet?
                > > I know it's impossible but why?
                >
                > It's not impossible.  It's just often difficult to do it
*right*.

                Is it that useful? It'd boost reliability, but would it
boost performance? If
                you've got a card that can keep up with the wire, then you
don't have any
                bottlenecks that having two cards would solve.

                --
        
/==============================================================\
                |John Stracke    | http://www.ecal.com |My opinions are my
own.|
                |Chief Scientist
|=============================================|
                |eCal Corp.      |Help stamp out vi in our lifetime!
|
                |francis(_at_)ecal(_dot_)com|
|
        
\==============================================================/

                



<Prev in Thread] Current Thread [Next in Thread>