Cisco has introduced Nexus. Nexus are the new line of data center switches. They come in the variables of 7000, 5000, 2000, and lastly 1000.
Nexus 7000 with their functionality sit at distribution layer, while Nexus 5000 come in the access layer. Nexus 2000 are nothing but extension to the 5k switches. In easier analogy, they work as line cards in 6500 chassis.
In this article, I would say why or when to use Nexus 7000 in the enterprise core layer. By purpose, the Nexus was designed for data center. But with the increased requirements of backbone network, and network growth the current top of the line 6500 switches comes short.
The backplane/fabric of 6500 switch is part of Supervisor Engine, in the case of SUP720 a 40GB per line module is the maximum bandwidth. What happens if you connect 8 ports with speed of 10G line card is that we are oversubscribing 1:2 ratio. This will be doubled with 16 ports of 10 GB line card to 1:4. The issue when multiple (30 or more) distribution switches are linked with 10G then the chassis with 9 slots becomes not enough. That is in the case of connecting without oversubscription.
The Nexus switches have different architecture. They line cards don’t depend on Supervisor Engine’s fabric for traffic processing exclusively. Each module will have its own fabric. This fabric rather connected by a fabric module that can be upgraded by itself. Each fabric supports 46GB per slot. Nexus 7000 with 10 slots support 5 fabrics, that is equal of 230GB per module slot. This is 5.75 times more than the original 6500 fabric. Still, a 32 10GIG port have a 80GP backplan, thus these 32 ports line card are oversubscripted at the rate of 1:4. while the 8 port Line cards are Non-oversubscripting.
Honestly, I still can’t figure out the reason for their fabric modules, since their line card modules are all having a limit 80GP fabric. So with 8 line cards, we have requirement of 640 GP, and we still lacking the support from the fabric module for these line card.