CCNP Switch: Punt vs Glean

Node Man

Hi Everybody,
Can anyone explain the difference between Punt and Glean? The more I look at it, the less sense it makes.

Thanks!

joetest

*This is based off of memory*
Punt adjacency
Punted packets are packets that can't be CEF switched and has to be process-switched by the CPU. It happens if it's a packet which needs fragmentation(think MTU), is encrypted or if the router/switch itself is the destination. Packets going through NAT or ARP packets is also punted packets. Show the number of punted packets with: show ip cef switching - old command was show cef not-cef-switched.
You can find Punt adjacencies using: show ip cef adjacency punt

Glean adjacency is when your device has multiple hosts connected via the same interface(ethernet ie.). Instead of having an entry for each specific host in the adjacency table you have the specific prefix they reside in instead.:


Makes sense?!

Node Man

Thanks Joe. It sounds like Glean creates route auto-summaries?

fredrikjj

Node Man wrote: »

Thanks Joe. It sounds like Glean creates route auto-summaries?

No, it's about minimizing the amount of information you need to store about a multiaccess interface in the adjacency table. Let's say that you have an Ethernet link between routers R1 and R2. You configure the interfaces with 10.1.12.1/24 on R1 and 10.1.12.2/24 on R2. However, a router doesn't automatically know what IP address that you have configured on the other side of the link. On R1, R2's IP address could be learned as a next hop through a routing protocol. To send a packet to R2, R1 needs the MAC address that belongs to the next hop IP address, information it gets through ARP (or NDP if IPv6). The IP to MAC mapping information is stored in the CEF structure called the adjacency table.

The information from the routing table (prefix and next hop) and the mapping information from the adjacency table is used to forward the packet. CEF on R1 may or may not know all next hops that exists on the link and have them resolved to a hardware address; it only knows about .2 because a routing protocol has told R1 about that next hop and knows the MAC address because it has been resolved by ARP.

On R1:

FIB:
I know about the prefix X and the next hop is 10.1.12.2.

Adjacency table:
I know the MAC address that maps to 10.1.12.2
I know that other destinations in the 10.1.12.0/24 may need to be resolved in the future <--- this is the "glean adjacency"

At any point, another router on that link, R3, could inform R1 about the 10.1.12.3 next hop, and if packets need to be sent to that IP address, new mapping information has to be acquired. Though, if you have routing protocol running, the next hop would be resolved before actual data plane traffic is sent, but you could imagine a static route instead that would be installed in the FIB without there being a mapping the adjacency table, and data plane traffic would hit the glean adjacency which would trigger next hop resolution.