design using muliple B-port FDDI attachments ?
James Kovaly
We are connecting three of our sites to a non-shared FDDI ring via alternate
Southern Bell COs. Each CO will contain an ODS fiber hub where we will
connect to an M-port on the hub. The hubs will be A-B connected to other hubs
within Southern Bell. Two fiber pairs will be available at each of our
sites. The Protocol will be TCP/IP.
Each of our sites is setup with Wellfleet routers. Since downtime is not
tolerated I am suggesting that instead of standard dual-homing, one of the
fiber pairs to an A-port and the other pair to the B-port of a single FDDI
card on the Wellfleet router to the M-ports on the ODS hubs at the COs,
that we opt to setup each of the Wellfleet routers with (2) FDDI cards and
connect a B-port on each of those cards to the M-ports on the ODS hubs,.
Using a higher RIP cost on one of the B-ports or possibly OSPF (for load
balancing), the failure of an FDDI card will not disrupt service (except maybe
for the short route re-convergence time). Dual homing with the A - B port
method would only provide redundancy in case of a link, CO, hub or FDDI port
failure and not in case of an FDDI card failure in the router.
I realize that using RIP would necessitate the use of two IP networks on
the same FDDI ring but the routers should be able to handle this and I don't
believe this to be a problem for Southern Bell since they have other FDDI
rings that provide shared service to some customers.
Thanks,
---------------------------------------------
James Kovaly
Network Design Engineer
BellSouth Cellular Corp.
---------------------------------------------
Brett Frankenberger
You'll need two IP networks on the same ring, regardless of what
Routing Protocls (RIP, Static Routes, etc) you use.
It will cause nasty loops, but you can filter to avoid it. Condier the
case of two two subnets on the FDDI: 1.1.1.0 and 1.1.2.0 (both with
255.255.255.0 masks). Assume the first router always uses host adress
1 (so it has 1.1.1.1 on FDDI #1 and 1.1.2.1 on FDDI #2), the second
router uses host address 2, and the third uses 3.
Now, suppose a user types ping 1.1.1.255. Your network just died.
(Well, not really, but it will generate a ton of packets on your FDDI
... and if enough people do it, it can kill your network). Why? LEt's
assume the user is on another (say, Ethernet) port on router #1.
Router #1 receives the packet, and knows it is a broadcast for the
1.1.1.0 network, and it knows the 1.1.1.0 network is locally attached.
So it sends the packet out on the FDDI ring as a broadcast. This will,
of course, be received by the 1.1.1.2 and 1.1.1.3 interfaces. But the
router will ignore those - they are broadcasts received from the
networek they are for, so they are ignored. But, interfaces 1.1.2.1,
1.1.2.2, and 1.1.2.3 will also receive the broadcast. And the router
will forward them to the 1.1.1.0 subnet (since that's what routers do -
they route packets). They have no way of knowing that it's actually
being put back out on the same physicaly network they can in from. So,
now you have three copies of the frame on the network. Those three
will loop, so you net nine more. Then 27 more. And so on, until the
TTL expries. 32 is a fairly common TTL in the IP world, so you end up
with 308 Trillion Packets! (Obviously not ... the router will drop
most of those because of no buffers, but the point is that it will
briefly bring down the ring).
In short, whenever you have multiple subnets on a ring, you need to
define multiple IP addresses on the routers on that ring so that every
router interface on the ring had an address on every subnet on the
ring. But, you cannot do this. (The whole point is to put one subnet
on one router interface and a different subnet on the other).
So, you need to write filters to deal with the broadcast packets.
Have fun :)
--
- Brett (bre...@netcom.com)
------------------------------------------------------------------------------
... Coming soon to a | Brett Frankenberger
.sig near you ... a Humorous Quote ... | bre...@netcom.com