markat@cyrano.llnl.gov (Mark Tassinari) (03/05/91)
Greetings.
We just purchased a number of cisco AGS+ routers and intend to begin routing
IP, DECnet and eventually Appletalk. I have heard somewhere that sites are
configuring their nets to take advantage of cisco's high bandwidth backplane.
We were hoping to do this also, but are running into some problems with regard
to IP addressing. I hope someone that has traveled this road can offer some
advice on how to set up the IP addressing.
The concept was simple. Rather than having a 10Mb ethernet connect all the
node routers together we would connect them to a single cisco and use the
backplane as the backbone.
Bld1 Bld2 Bld3
\ | /
\ | /
--------- --------- --------- ---------
| Node1 | | Node2 | | Node3 | | Node4 |
| cisco | | cisco | | cisco | | cisco |
--------- --------- --------- ---------
\ \ / /
\ \ / /
\ \ / /
\ \ / /
\ \ / /
-------------
| hub |
| cisco |
-------------
Each of the Bldx networks is assigned one class B subnet. My big question is,
what's the best way to address the node<->hub links? Its seems wasteful to
use a class B subnet for each. We tried mucking with the netmask to make a
single subnet into a number of smaller subnets and using one of these small
subnets on each link. We then got into trouble with conflicting netmasks.
Is there a better way?
Thanks.
~~~~~~~~~~~~~
Mark Tassinari markat@cyrano.llnl.gov
Lawrence Livermore National Lab (415) 422-4090
PO Box 808 L-408 Livermore, CA 94550haas%basset.utah.edu@cs.utah.edu (Walt Haas) (03/05/91)
In article <32870@boulder.Colorado.EDU> markat@cyrano.llnl.gov (Mark Tassinari) writes: >We just purchased a number of cisco AGS+ routers and intend to begin routing >IP, DECnet and eventually Appletalk... Rather than having a 10Mb ethernet >connect all the node routers together we would connect them to a single >cisco and use the backplane as the backbone.... What do you hope to gain by using the extra [hub] router? We use the usual backbone architecture with an Ethernet backbone (to be upgraded to FDDI). I can't see what would be gained by replacing the Ethernet with a router. -- Walt
kwe@bu-it.bu.edu (Kent England) (03/05/91)
> From: markat@cyrano.llnl.gov (Mark Tassinari) > Subject: Network config questions > > Bld1 Bld2 Bld3 > \ | / > \ | / > --------- --------- --------- --------- > | Node1 | | Node2 | | Node3 | | Node4 | > | cisco | | cisco | | cisco | | cisco | > --------- --------- --------- --------- > \ \ / / > \ \ / / > \ \ / / > \ \ / / > \ \ / / > ------------- > | hub | > | cisco | > ------------- > > > My big question is, > what's the best way to address the node<->hub links? Walt Haas asked the right question: Why the hub cisco if you have leaf node ciscos? I ask the complementary question: Why leaf node ciscos if you have the hub node cisco? If what you want is to use the cBus as your high speed backbone, I suggest you plug each interface of the hub cisco directly onto each subnet in each building, using fiber optic ethernet. If you have more than 18 or so Ethernets you need more than one AGS+ chassis and then you have to link the hub routers together with something. Perhaps FDDI in future. This approach does away with the need for two port subnets. However, whenever it is that cisco supports OSPF then you will have variable length subnet masks and link state routing, both big wins. With variable length subnets, you don't run into the problem you described with subnet mask conflicts. Your hosts are not affected and can still use a common subnet mask and you can conserve subnet space for lots of small subnets. There are two ways to plug the hub cisco interfaces into each subnet: you can use the ODS fiber optic transceiver cable extender solution or you can use a pair of standard FOTs. The ODS solution requires an Ethernet cable to plug into on the subnet side, which is an advantage if you don't have a repeater or bridge to attach to. The "pair of FOTs" solution is best if what you are attaching to is the AUI port of a multiport concentrator. FOTs give you better link level error detection than does the ODS solution which is pretending to be an AUI cable. If anyone knows of another AUI cable extender vendor than ODS, let me know and I'll mention them in future. All things equal, I prefer FOTs, but I invite contrary comments and experience. We have had some trouble with ODS units aging, but we don't have a lot of experience yet with FOTs. Cabletron is our preferred vendor today for FOTs. --Kent
robel2@mythos.ucs.indiana.edu (Allen Robel) (03/18/91)
I'll suggest the following modification to your diagram
so as not to waste a subnet:
Bld1 Bld2 Bld3
\ | /
\ | /---- B
--------- | --------- --------- ---------
| Node1 | | | Node2 | | Node3 | | Node4 |
| cisco | | | cisco | | cisco | | cisco |
--------- | --------- --------- ---------
\
\
\
\
A \
-------------
| hub |
| cisco |
-------------
Thus, the link to the node cisco is also the link to
one of the buildings. One way to implement B in the above
diagram, if you're using fiber, is to use a multiport
fiber repeater with one port feeding Node1 and the other Bld3.
If you're worried about the cost of the repeater, consider that
you're saving an interface on the cisco. In detail, it would
look like...
Bld 1
|---[fiber xcvr]-----[fiber repeater]
[Hub cisco][fiber xcvr]----|
|---[fiber xcvr][Node1 cisco]
MultiPort
Fiber
Repeater
A disadvantage is that you'll have another repeater in
the signal path to Bld3 so will need to watch the number
of stacked repeaters (10baseT hubs etc.) in that building.
Another advantage, though, is that Bldg3 gets fewer hops to
the hub cisco (so let it be the one in which the Provost or
other Big Cheese resides :-)
regards,
Allen Robel robelr@mythos.ucs.indiana.edu
University Computing Services ROBELR@IUJADE.BITNET
Network Research & Planning voice: (812)855-7171
Indiana University FAX: (812)855-8299robel2@mythos.ucs.indiana.edu (Allen Robel) (03/18/91)
I'm not sure if this made it to the list so am resending
(with a correction to the second diagram...)
I'll suggest the following modification to your diagram
so as not to waste a subnet:
Bld1 Bld2 Bld3
\ | /
\ | /---- B
--------- | --------- --------- ---------
| Node1 | | | Node2 | | Node3 | | Node4 |
| cisco | | | cisco | | cisco | | cisco |
--------- | --------- --------- ---------
\
\
\
\
A \
-------------
| hub |
| cisco |
-------------
Thus, the link to the node cisco is also the link to
one of the buildings. One way to implement B in the above
diagram, if you're using fiber, is to use a multiport
fiber repeater with one port feeding Node1 and the other Bld3.
If you're worried about the cost of the repeater, consider that
you're saving an interface on the cisco. In detail, it would
look like...
A B Bld3
|-|-----------[fiber repeater]
[Hub cisco][fiber xcvr]----| |
|-|---[fiber xcvr][Node1 cisco]
MultiPort
Fiber
Repeater
A disadvantage is that you'll have another repeater in
the signal path to Bld3 so will need to watch the number
of stacked repeaters (10baseT hubs etc.) in that building.
Another advantage, though, is that Bldg3 gets fewer hops to
the hub cisco (so let it be the one in which the Provost or
other Big Cheese resides :-)
regards,
Allen Robel robelr@mythos.ucs.indiana.edu
University Computing Services ROBELR@IUJADE.BITNET
Network Research & Planning voice: (812)855-7171
Indiana University FAX: (812)855-8299Allen Robel <robelr@mythos.ucs.indiana.edu> (03/21/91)
> Bld1 Bld2 Bld3 > \ | / > \ | /---- B > --------- | --------- --------- --------- > | Node1 | | | Node2 | | Node3 | | Node4 | > | cisco | | | cisco | | cisco | | cisco | > --------- | --------- --------- --------- > \ > \ > \ > \ > A \ > ------------- > | hub | > | cisco | > ------------- Another drawback to my suggestion is that traffic from Bld1 & 2 is repeated onto Bld1 on its way to the node cisco. A bridge in Bld3 would solve this but now things are looking kind of jucky. :-( Allen Robel robelr@mythos.ucs.indiana.edu University Computing Services ROBELR@IUJADE.BITNET Network Research & Planning voice: (812)855-7171 Indiana University FAX: (812)855-8299