Vlan routing questions on the 6509

Question

   We have a 6509 VSS at our main site and one vlan (an IP class C size) is comprised of a large number of servers with single Gb interfaces. These are connected to the 6509 via various methods - blade centers with GB portchannels, some directly attached to the core, and some via 4948s with 10Gb trunk uplinks. My question is this...I know we have way too many servers in one subnet (this is not all of our servers) and I know that all broadcasts will hit every individual server but how does the 6509 ASICs handle the packets in and out of the vlan with multiple connections to that vlan on the 6509s?  Can packets get routed through that 6509 vlan router interface simultaneously from the multiple layer 2 connections on the 6509s? What I am asking is if the 6509 vlan routing interface throttles all the Gb interfaces into a single GB interface through the vlan routing interface? What about the few servers on the 10Gb interfaces - are they throttled to a single routing 10Gb interface or does each connection have it's own connection to the routing interface?  What I want to know is if the 6509 acts as though it was like a single separate (1G and 10Gb) router attached to the vlan  - like a bunch of switches connected together with a single separate router attached to one of the interfaces for routing out of the vlan. Thanks

Answer 1

Packet switching within the vlan is not what I am asking about. I want to understand the process the 6509s use when they route from a vlan (one subnet) to another vlan (subnet) - L3 routing out of the vlan.
It's pretty much the same thing which is what Reza was explaining.
It can be helpful sometimes in terms of design etc. to think of a L3 switch as you would if it was a physical router and L2 switches but in terms of forwarding thinking of it like that is misleading.
In terms of forwarding L3 traffic the SVI does not correspond to the physical interface of the router. The actual interfaces used would, in the case of the 6500, be the physical port connections for the source and destination devices on their corresponding linecards.
If the linecards did not have DFCs then a part of the packet is sent by the linecard to the PFC for a forwarding decision. If they do have DFCs then they can make the forwarding decision locally.
Either way the forwarding decision is made by looking at the FIB (Forwarding Information Base) which is stored on the PFC and on each DFC if the linecards have them. The FIB should have entries for connected and remote networks (learnt via the IP routing table), the next hop IP and it's L2 mac address so all the information needed to forward the packet at L3 is there.
So, as Reza says, the packet is then switched either locally on the linecard from one port to another or is sent from the ingress linecard to the linecard with the egress port via the switch fabric.
Any bottlenecks within the chassis apply to both L2 and L3 forwarding eg oversubscription etc.
The above is a very high level view of how it works. If you want to understand it in greater detail it would be worth having a read of the link Reza provided.
Jon