cef punt..

why?

kick it...

the fib is full (i'm all outta fib)
entry can't be located by the fib
ttl has expired
fragment required because mtu has been exceeded
icmp redirect needed
unsupported encapsulation
tunneled packets needing encryption or compression
acl with log operation
nat needed, except for 6500 sup 720 or better which can do nat in hardware

ask fibber...

MLS synchronicity...

simply cef...

route caching is comprised of a l3 process and a layer 2 engine

the route process ships a flow's first packet, then the l2 process takes control by shipping the rest of the flow
a flow is a unidirectional stream of packets that share the same protocol... if this is a joint communication then naturally, there are two flows being managed

the downside of route caching is the delay caused by software processing  the first packet

enter CEF, hardware based switching, faster as a result and requires less processing power

CEF keeps two tables a FIB (forward information base) and an adjacency table.  the fib is basically the routing table formatted differently

as ARP discovers new  and adjacent hosts, this information is updated in the adjacency table

with MLS,  four tables are now used for packet switching...

cef'ing off...

cef provides:

performance - cef is less cpu intensive than fast switching or route caching, giving way to processing for other layer 3 services such as qos

scalability - cef enables full switching capacity for line cards, however it is also used on fixed configuration switches

resilience - because it uses fib lookups which contain all known route destinations; fast switch processing and route cache dependance are eliminated

the primary components of CEF are the forward information base and adjacency tables

adjacency table maintains layer 2 linked to a particular fib... this avoids arp reliance

types of adjacency

cache adjacency - this entry contains the outbound interface and the corresponding mac address for its fib entry.  the mac is the destination subnet that is directly connected, or the mac of the subnet the packet must be forwarded to to reach the destination

receive adjacency - this entry handles packets whose final destination is THIS router.  includes packets assigned to the router, broadcast packets, and multicasts which have as their destination THIS router

null adjacency - handles null interface packets; these packets are usually dropped

punt adjacency - packets that cannot be handled by cef, usually forwarded to fast switching layer for forwarding

glean adjacency - created when when the router knows the directly connected destination ip subnet and does NOT know the destination mac, or knows the forwarding router's destination subnet but does NOT know that router's mac address.  these packets generate an arp request

discard adjacency - fib entries to these adjacencies are discarded

drop adjacency - packets to this entry are dropped but the prefix is checked

dCEF or distributed cef is recommended when there is a fib on each line card in a chassis.  this precludes the main processor and/or route table queries for next hop ip's.  fast switching is instead performed

fools rush in where angels fear to tread...

a router is a switch, at least in the mls universe...
or a router and switch combined...
or is the switch a router?

these prattlings of mine, are just that, mine... they have their basis in fact but they are sifted through me, and i am a fallible instrument of their conveyance... as always, question everything and  prove the truth for yourself...
in the end the numbers will out...

the l3 switch has the properties of both the router and the switch... it maintains the route table, the arp table, the fib and the adjacency table, among others... cef is built in... the last two tables are considered distinct functional blocks... they are functionally different but make up the whole... they are organized by occurences to the route table and arp table, as we saw in the last submission, and are functionally independant as well as functionally dependant... all four are tables maintained by the mls...

what happens when you show the mac-address-table of a router?


r2620_01#sh mac-add

r2620_01#

nothing...

and the ip route table of a switch...


sw2950_02#sh ip route
                 ^
% Invalid input detected at '^' marker.

sw2950_02#

disaster...
however...


sw3550_01#sh ip route

trimmed

     1.0.0.0/24 is subnetted, 1 subnets

D       1.1.1.0 [90/130816] via 192.168.1.50, 2d16h, Vlan10

                [90/130816] via 172.16.50.50, 2d16h, Vlan50

     2.0.0.0/24 is subnetted, 1 subnets

D       2.2.2.0 [90/2298112] via 192.168.1.50, 2d16h, Vlan10

                [90/2298112] via 172.16.50.50, 2d16h, Vlan50

     100.0.0.0/8 is variably subnetted, 2 subnets, 2 masks

C       100.0.0.0/24 is directly connected, Loopback0

and...


sw3550_01#sh mac-add
          Mac Address Table
-------------------------------------------

Vlan    Mac Address       Type        Ports
----    -----------       --------    -----
 All    000f.8ffe.0980    STATIC      CPU
 All    000f.8ffe.0981    STATIC      CPU
trimmed


the two worlds collide beautifully; synchronicity is achieved...

a diagram that depicts distinct models of a router with a route and arp table, and a separate model of  the mls with fib and adjacency, and the interaction of the two as if they are completely separate entities is misleading...

the router is a switch... the model is a layer 3 switch and it maintains a route table, an arp table, a fib and an adjacency table, among others...

a cef is born...

route one switch many...

in the beginning multi-layer switching began as a marriage between a route processor and a switch engine and it was good.  the rp gets the first packet, determines a route and forwards the packet to the correct interface.  the se  participates by learning this path from the rp, and if it is determined that the switch engine can switch both directions, the se cuts out the routing process.  this is also called route cache switching or netflow switching...

cisco express forwarding evolved from netflow switching and is more efficient because it uses built in hardware and dynamic lookups... cef's first functional block is: the fib, forwarding information base...

the FIB is not a lie
the fib block is the layer 3 engine, the router.  to create the fib, the router builds a  table and orders the most specific route to the destination first, the longer the destination prefix, the closer to first place.  the fib contains the next hop entry for each destination.  the fib also contains a host entry, or the subnet mask of the address in all 1's or 255.255.255.255. after a topology change the fib is updated by the router, and if an arp entry ages out, the fib will again be updated to reflect that change.


sw3550_01#sh arp
Protocol  Address          Age (min)  Hardware Addr   Type   Interface
Internet  192.168.1.111           0   0009.b752.d780  ARPA   Vlan10
Internet  192.168.1.100           -   000f.8ffe.0980  ARPA   Vlan10
Internet  192.168.1.120           0   0014.6ad0.da00  ARPA   Vlan10
Internet  192.168.1.112           0   0009.b73f.ce80  ARPA   Vlan10
Internet  192.168.1.50            0   000f.2394.6c40  ARPA   Vlan10
Internet  192.168.1.1             0   c03f.0eab.d1ec  ARPA   Vlan10
Internet  192.168.69.69           -   000f.8ffe.0980  ARPA   Vlan69
Internet  172.16.50.2             0   000d.ed22.eb44  ARPA   Vlan50
Internet  172.16.50.50            0   000f.2394.6c40  ARPA   Vlan50
Internet  172.16.50.100           -   000f.8ffe.0980  ARPA   Vlan50
Internet  192.168.1.130           0   0011.5c2d.b8c0  ARPA   Vlan10

of course the command is not show ip fib, rather...

sw3550_01#sh ip cef
Prefix               Next Hop             Interface
0.0.0.0/0            192.168.1.1          Vlan10
0.0.0.0/32           receive
1.1.1.0/24           172.16.50.50         Vlan50


omitted
192.168.1.130/32     attached             Vlan10
192.168.1.255/32     receive              Vlan10
192.168.69.0/24      attached             Vlan69
192.168.69.0/32      receive              Vlan69
192.168.69.69/32     receive              Vlan69
192.168.69.255/32    receive              Vlan69


and after the vlan information has been eliminated:
no more arp for 192.168.69.69...

sw3550_01(config)#do sh arp
Protocol  Address          Age (min)  Hardware Addr   Type   Interface
Internet  192.168.1.111           1   0009.b752.d780  ARPA   Vlan10
Internet  192.168.1.100           -   000f.8ffe.0980  ARPA   Vlan10
Internet  192.168.1.120           1   0014.6ad0.da00  ARPA   Vlan10
Internet  192.168.1.112           1   0009.b73f.ce80  ARPA   Vlan10
Internet  192.168.1.50            1   000f.2394.6c40  ARPA   Vlan10
Internet  192.168.1.1             0   c03f.0eab.d1ec  ARPA   Vlan10
Internet  172.16.50.2             1   000d.ed22.eb44  ARPA   Vlan50
Internet  172.16.50.50            1   000f.2394.6c40  ARPA   Vlan50
Internet  172.16.50.100           -   000f.8ffe.0980  ARPA   Vlan50
Internet  192.168.1.130           1   0011.5c2d.b8c0  ARPA   Vlan10

and  the fib entry is also eliminated...

sw3550_01(config)#do sh ip cef
Prefix               Next Hop             Interface
0.0.0.0/0            192.168.1.1          Vlan10


omitted
192.168.1.130/32     attached             Vlan10
192.168.1.255/32     receive              Vlan10
224.0.0.0/4          drop
224.0.0.0/24         receive

or specifically...

sw3550_01(config)#do sh ip cef vlan 69

sw3550_01(config)#