Our sample application performs IPv4 five-tuple exact-match classification on the IP packets extracted from the Ethernet packets. The microblock that implements this classification takes the packet and modifies two pieces of packet metadata based on the contents of the packet: the flow ID, which in our application also identifies the ring number on which to put the packet after the packet has been processed, and a next-hop ID identifying the next hop to which the packet should be forwarded. This classifier is more useful as a programming example than it would be in the real world for a couple of reasons. First, this classifier forwards packets to next hops based on an exact match of the fivetuple, which is not normally how packets are forwarded. Second, this classifier ignores some aspects of IP packets that probably should not be ignored in a production environment, such as the handling of packets with IP options and ICMP packets. These aspects make the classifier a simple example for the purposes of this book. A more complicated, industrial-strength classifier can be found in the Intel IXA SDK 3.0 reference designs.

The IPv4 five-tuple consists of five fields in the packet: IP source and destination addresses, IP protocol number, and the next-layer source and destination ports. The IP protocol number should not be confused with the Ethernet protocol number. The Ethernet protocol number defines the type of packet inside the Ethernet packet, while the IP protocol number defines the type of packet...