[c-nsp] router-switch redundancy

[Jon Lewis]

> On 8/13/05, Mark Kent <mark at noc.mainstreet.net> wrote:
>> In more than a few places on the cisco web site they have
>> diagrams like this:
>>
>>            {Internet}
>>             /      \
>>            /        \
>>        [routerA]   [routerB]
>>         /   \       /    \
>>        /     \     /      \
>>       /       \   /        \
>>      /         \ /          \
>>     /           /            \
>>    /           / \            \
>>   /           /   \            \
>>  -------------    ---------------
>>  |           |    |             |
>>  |  switch   |----|   switch    |
>>  |           |    |             |
>>  -------------    ---------------
>>
>> designed so that either router can fail and things still work
>> for the servers connected to the switches.
>>
>> What I don't get about this is the two connections
>> from each router into the same switch fabric.

Is it really necessary?  I did a setup similar to this in our data center 
and each "router" has just one connection to the switch fabric, and the 
routers are directly connected via fiber.  If either router or either of 
the main switches fails, the network still works.

So my setup is more like:

             {Internet}
              /       \
             /         \
        [swouterA]---[swouterB]
           |             |
   -------------    ---------------
   |  switch   |----|   switch    |
   -------------    ---------------

----------------------------------------------------------------------
  Jon Lewis                   |  I route
  Senior Network Engineer     |  therefore you are
  Atlantic Net                | 
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________