[comp.protocols.tcp-ip] cisco question

AFDDN.TCP-IP@GUNTER-ADAM.AF.MIL (10/11/89)

Anyone know the answer to this one???

  Within the DDN Milnet, we are running IP over X.25.  GENERALLY speaking,
when a host has to send an IP datagram, it looks at the IP address of the
destination (or the next hop), then claculates the X.121 style address using
the DDN's algorithm.  If there's already a virtual circuit established to
that destination, then the datagram is queued and sent on that vc.  If there's
no vc already established, then the call request procedure is initaited and
a virtual circuit to the correct destination comes into being.  The end result 
of this behavior is that only one X.25 vc is established between any given
pair of IP addresses on the same network (what a concept).

  Now for the meat of the question.  Let's suppose that I have my own packet
switch with 20 ports on it.  I have a cisco MGS or AGS gateway and I connect
my packet switch to the Milnet via the gateway.  I get the NIC to assign me
a class B net number such as 131.7.0.0.  I could then assign each of my 20
ports to have a different number in the forth octet, so I would have 
something like 131.7.x.1 thru 131.7.x.20.  I could also decide to connect
lans to my packet switch and subnet my class B addresses.  For instance, I
could let 131.7.1.x be a subnet that includes the cisco and each of the ports
which would give me 131.7.1.1 thru 131.7.1.254, of which I only need 20 
because I only have 20 ports.  Now, if I decide to connect a gateway from a
lan to my packet swicth on port 5, its address in subnet 131.7.1 would be 
131.7.1.5.  I could let the lan be subnet 131.7.32.x thru 131.7.255.x.  I
assume I could break the third octet on any convenient boundary. 

  now my problem is how to tell the cisco how to route to the gateway to
the lan based on the subnet field being in a range.  Any one see how this
would be done??  The second part of the question (assuming it can be done),
is whether the cisco will only open one vc to the lan gateway or 
whether it will try to open a vc for each unique IP destination address on 
the lan??

darrel b.
-------

satz@CISCO.COM (Greg Satz) (10/11/89)

>>   Now for the meat of the question.  Let's suppose that I have my own packet
>> switch with 20 ports on it.  I have a cisco MGS or AGS gateway and I connect
>> my packet switch to the Milnet via the gateway.  I get the NIC to assign me
>> a class B net number such as 131.7.0.0.  I could then assign each of my 20
>> ports to have a different number in the forth octet, so I would have 
>> something like 131.7.x.1 thru 131.7.x.20.  I could also decide to connect
>> lans to my packet switch and subnet my class B addresses.  For instance, I
>> could let 131.7.1.x be a subnet that includes the cisco and each of the ports
>> which would give me 131.7.1.1 thru 131.7.1.254, of which I only need 20 
>> because I only have 20 ports.  Now, if I decide to connect a gateway from a
>> lan to my packet swicth on port 5, its address in subnet 131.7.1 would be 
>> 131.7.1.5.  I could let the lan be subnet 131.7.32.x thru 131.7.255.x.  I
>> assume I could break the third octet on any convenient boundary. 

Is the following picture equivalent to the above words?

    ---26---ciscoM---131.7.1.1---[switch]---131.7.1.5---ciscoL---131.7.x.y
      X.25             X.25                   X.25                LAN

>>   now my problem is how to tell the cisco how to route to the gateway to
>> the lan based on the subnet field being in a range.  Any one see how this
>> would be done??  The second part of the question (assuming it can be done),
>> is whether the cisco will only open one vc to the lan gateway or 
>> whether it will try to open a vc for each unique IP destination address on 
>> the lan??

The way to make this work is to inform the routers of each other's
presence. This can be done using static routing or with a dynamic routing
protocol such as IGRP. The cisco connected to Milnet, ciscoM, must know
that everything on 131.7.0.0 is reachable out its X.25 interface connected
to your 20 port switch. The cisco connected to the LAN, ciscoL, must know
that the rest of the world is accessible via its X.25 interface.

Whether you use static or dynamic routing, the routing database will
contain the next hop address which will either be 131.7.1.1 for ciscoL or
131.7.1.5 for ciscoM. As these two routers get involved in all traffic
between them, a minimum of one VC is used. More can be permitted using the
X.25 NVC parameter.

For example, if ciscoM needed to send something to 131.7.32.1, it would
find that 131.7.5.1 was the next hop address. Furthermore if ciscoL needed
to send something to 10.2.0.2, it would find the next hop to network 10 is
131.7.1.1. In both these cases, your excellent description holds.

For more configuration details, please send mail to
customer-service@cisco.com.

Greg Satz
cisco