11-24-2010 07:23 AM - edited 03-06-2019 02:11 PM
today we discussed an interesting situation: assume we have a Nexus 7000 equipped with two M- and one F-linecards. Hosts in VLAN 100 and VLAN 200 are connected to the F-linecard only. Now hosts from VLAN 100 need to communicate with hosts in VLAN 200 - and vice versa. So the routing must be done by SVIs for VLAN 100 and VLAN 200 handled on the M-linecards. Which of the two M-linecards is now doing inter-vlan routing?
11-24-2010 09:46 AM
Evenin'
That is a good question.
In short, both cards. Without going into too much detail, the various flows will be load balanced accross each of the M1 line cards, similar in principle to how traffic is load balanced on a port-channel hash.
Chris
11-24-2010 11:16 PM
Chris,
thanks for your feedback... this means also: since the m-linecards have a 80Gbps fabric-connaction and we are load-balancing accross two of them we would achieve a L3 performance of 160Gbps - right? If yes, does this concept scale higher - something like 4*M = 320 Gbps L3 performance - or x*M = x*80 Gbps L3 performance?
Cheers,
/Peter
11-26-2010 01:09 AM
Peter,
Your understanding is indeed correct, although the load sharing is as important to spread the load over the forwarding engines as it is over the fabric (for scenarios where you have a heavy amount of local switching).
I do not have the functional specification in front of me, but based on the M1 architecture I would expect it to scale to 8 M1 cards. Although keep in mind that these M1 line cards would be shared between all F1 cards, rather than dedicated to specific modules.
Chris
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide