[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: subnets & 2 NICS in a mashine



On Friday 10 November 2000 08:27, robert_wilhelm_land wrote:
> Robert Guthrie wrote:
... An analogy that illustrated a wrong concept...

> Exactly what I assumed.
>

... and another bad analogy illustrating what really does happen on a single 
network (no gateway involved).

>
> I'm not to sure if this is a bit to simple.
> What I had drawn out of some articles is:
> When data is supposed to be sent to a certain computer and this
> computer is not 'online' or wthin the subnet, the data would be
> adressed to the gateway mashine and let this then decide what to do.
> In case of the adressed mashine saying "here I am!" (because of being
> in the same subnet) the data would be chopped in IP packets with each
> packet containing a header with the IP-number or MAC number to be
> adressed correctly. Otherwise - I guess - would the data be bounced
> throughout the network and cause useless traffic?
>
Yes, when the destination for a packet is not on the local subnet, one of the 
router/gatways connected to that subnet will then forward that packet to the 
next subnet.  But if the destination is on the same subnet as the original 
sender, then that sender never recieve an acknolegement of recieving that 
packet; some programs will continue to re-send tha package until a certain 
time passes without any responses.  

The gateway NEVER does anything with a packet that is destined for the same 
subnet that it started on.  It decides if it is to be forwarded by using the 
netmask.  Here an example will be helpfull:

Netmask for 192.168.2.0 network: 255.255.255.0
Computer A's Source IP address: 192.168.1.20
Gateway (computer G) IP addresses: 192.168.1.1, and 10.110.1.1

If A sends a packet to 192.168.2.45
Everyone on the system sees this packet, including the gateway and the 
destination computer.  The destination computer sends an aknowledgement back 
to A.  The gateway, not paying attention to this conversation, masks 
192.168.2.45 against 255.255.255.0 (if you know binary math, this is a 
logical AND of all the bits), which yields 192.168.2.0, which is equal to the 
network address mentioned above.  Because the masked destination is equal to 
the network address, the gatway does not forward the packet.  THAT is how it 
decides if it moves the packet from one NIC to another.  

If A sends a packet to the desination of 10.110.1.68, then the gatway masks 
that against 255.255.255.0, and gets 10.110.1.0, which is not equal to 
192.168.2.0, so it knows that it needs to forward it.


> Exactly this imagination makes me hard to understand why 2 NIC's on
> one mashine need two subnets. I recently configurated the two cards in
> the server to work within the same subnet which turned out that the
> w95 client could not adress anything to the smb server on the
> 2-NIC-mashine.

As you may have figured out, having 2 NICs on the same subnet would not do 
anything for you.  The computer with 2 NICs would either recieve duplicate 
packets, and have to do double the work (forwarding duplicate packets), or it 
would have to ignore one of the nics.  

It doesn't HAVE to be this way; someone could re-write the tcp/ip protocol if 
they wanted two NICs on the same subnet, but there's no point.  

---------------
One non-analogy my clear this up:  signals on copper wire are not 
directional; you can't send two packets at the same time in opposite 
directions because the network is designed as if there only one ciruit across 
the wire that connects all of a given physical network (not including 
routers, gateways, or anything that reads data off of one wire and puts it on 
another).  Don't think of signals as electric current, flowing from one point 
to another.  Signals are more like a light in the middle of the room, 
blinking on and off in morse code.

-- 
Did you know that if you play a Windows 2000 cd backwards, you 
will hear the voice of Satan?

That's nothing!  If you play it forward, it'll install Windows 2000.



Reply to: