1) the 6500 has a shared 32Gbps bus but can also provide dedicated connections per module. It provides this dedicated connection by use of the crossbar switching fabric.
For non-bloking part of question see answer 7)
2) The SFM was a module that could be used with a Sup2 to provide dedicated connections to the switch fabric per module. If you had a sup2 but didn't have an SFM then even if the module had the ability for dedicated connections to the switch fabric it couldn't use those connections, it had to use the shared 32Gbps bus.
With a sup2 the SFM was a separate module and went in slot 5/6 of a 6509 chassis. The sup720 has an inbuilt SFM so no need for a separate module. This is also why the sup720 goes into slot5/6.
3) MSFC is not the same as SFM. On the 6500 there is the control plane and the forwarding plane. The forwarding plane is concerned with forwarding the actual data and this is done usually by the PFC. The control plane is primarily responsible for peering with L3 neighbors and building the routing table. This is what the MSFC does.
Important to note that anything done by the PFC is in hardware, anything done by MSFC is in software.
4) With the right modules you can get up to 40Gbps per slot on a 6509 chassis. This is not the actual throughput as such. In terms of pps (packets per second) the figures can be looked up on the datasheet (apologies but a nbit busy today so you can do that yourself)
5) DFC is a daughter card that can be added to DFC capable modules. What the DFC does is allow local forwarding decisions to be made on the module itself without having to send the packet to the supervisor. Without a DFC, you have CFC (Centralised forwarding) where a part of the packet must be sent to the supervisor for a forwarding decision to be made.
DFC dramatically increases throughput on a 6500. Again the actual rates will be on the datasheet for the 6500.
6) sup32 only supports 32Gbps shared bus. 720 supports 32Gbps shared bus + dedicated connections to the switch fabric. Basically modules fall into one of 3 categories -
i) classic linecards - these only support connection to the 32Gbps shared bus
ii) fabric enabled linecards - these support dedicated connections to the switch fabric but also support connection to shared bus if the sup does not support dedicated connections to switch fabric
iii) fabric only linecards - these cards only support dedicated connections to the switch fabric ie. they cannot connect to the shared bus
so i) & ii) would both work fine with a Sup32 although ii) would only be connection to shared bus. iii) would not work with a sup32.
i), ii) & iii) will all work with a sup720.
7) oversubscription is simply when you have more bandwidth coming in on the physical ports than you have going to the switch fabric. Most modules on the 6500 are oversubscribed although by how much differs greatly eg.
WS-X6548-GE-TX is a 10/100/1000 ethernet module
WS-X6748-GE-TX is also a 10/100/1000 ethernet module
the 6548 has 1 dedicated connection to the switch fabric of 8Gbps.
the 6748 has 2 x 20 dedicated connections to the switch fabric
So in theory on the 6548 you could have 48 ports running at 1Gbps = 48Gbps but with only 8Gbps connection to switch fabric that is an oversubscription ratio of 48 / 8 = 6:1
with the 6748 you could have 48Gbps on the ports with 40Gbps connection to the switch fabric. Much lower oversubscription.
If you absolutely wanted to make sure there was no oversubscription you would only populate the 6548 with 8 ports leaving the other 40 empty. With the 6748 you could use 40 ports and leave 8 empty. In practice you don't really get all 48 ports running at 1Gbps concurrently but oversubscription could be a very real issue on the 6548.
Jon
