CEF and DCEF

[mautez_mah]

Hi

I am confusedd about the difference between CEF and DCEF :
(dCEF) mode – When dCEF is enabled, line cards maintain identical copies of the FIB and adjacency tables. The line cards can perform the express forwarding by themselves, relieving the main processor – Gigabit Route Processor (GRP) – of involvement in the switching operation.
CENTRAL CEF MODE – When CEF mode is enabled, the CEF FIB and adjacency tables reside on the route processor, and the route processor performs the express forwarding. You can use CEF mode when line cards are not available for CEF switching, or when you need to use features not compatible with distributed CEF switching.

Now if Central Cef resides on route processor and performs so CEF still rely on Router processor? but I know CEF is coming to slove issue that the router will forward based on H.W rather than Software? 
so here can any one expline plz 

[MHM Cisco World]

check this link it explain the different between CEF and dCEF.

[marce1000]

@MHM Cisco World  You forgot the link (tx)

 M.

[Jon Marshall]

It really depends on the device you are talking about. 

As an example on the 6500 switch the supervisor has an MSFC which is responsible for, among other things, building the FIB and adjacency tables and it has a PFC which is responsible for actually forwarding the packets. 

If the switch is using centralised forwarding mode then those tables are downloaded to the PFC on the supervisor so the packets are still being forwarded in hardware. 

Other models of switch and/or router may do things differently. 

Jon

[Joseph W. Doherty · Accepted solution]

As Jon mentions, the 6500 series was a good example.

On a 6500, all L3 routing might be done via the main supervisor card.  This was "normal" CEF.  Unfortunately, even the later 6500 supervisors didn't have a L3 PPS forwarding rating (e.g. sup720B [?] 30 MPPS [?]) that came anywhere near the volume of traffic that could be handled by high bandwidth capacity (data) line cards (i.e. 40 and/or [later] 80 Gbps per card).

To address this, often those line cards supported (often optional) DFC modules which would do the L3 forwarding in a "distributed" manner, hence "distributed" CEF.  Only with those DFC modules, could a 6500 begin to approach the needed PPS to run most of its high-bandwidth ports at wire-speed.  (NB: actually, I recall, even with the best supervisors and line cards with DFCs, the 6500 might fall short of PPS needed for all high-bandwidth ports at full bandwidth, concurrently, but, especially in the Enterprise, generally, 6500s never got loaded to such an extent.  [That noted, in an ISP environment, I did see 7600s, pretty much the same hardware, suffer "stress" issues when actually very highly loaded.])