Route Switch Technologies

Layer 3 switching is routing any way you look at it. Devices used to be more clearly defined—bridges and switches were Layer 2 hardware-based devices, and routers were Layer 3 devices that performed their operations in software. That is history. However, you may hear others talk about routing switches and switching routers.

Routing switches are more Layer 2–oriented with some upper-layer functionality. They use hardware to route, but generally don't run routing protocols. Switching routers are primarily Layer 3 devices that can also switch, and do run routing protocols. Either way, hardware application-specific integrated circuits (ASICs) are used for switching speed and performance. Routers are not as slow as they once were, so this is really a moot point.

Routing and switching are both very important concepts that allow the hierarchical design for campus networks. Many internetworking devices today not only provide rich Layer 2 and Layer 3 functionality, but also upper-layer features to allow for making security and quality of service (QoS) decisions.

Traditional routing uses destination-based packet forwarding according to the Layer 3 header addresses. The frame passes from hop to hop according to the best path, which is normally some function of bandwidth depending on the routing protocol. By adding a NetFlow Feature Card (NFFC) and enabling multilayer switching (MLS), the Cat5000 can shortcut the process and rewrite the frame header similar to the router. Just as Layer 3 devices shortcut on IP addresses, Layer 4 devices can shortcut on port values. Shortcuts at other upper layers are often referred to as Application Layer switching.

Cisco Express Forwarding (CEF)

Back in the Chapter 1, “Shooting Trouble,” the basics of routing and switching were covered. You reviewed how routers route to the destination network address and that they buffer and switch packets from the inbound interface to the outbound interface within the router. Performance is definitely affected by the switching type, but switching types have certainly improved over the years.

Fast switching (ip route-cache) has been the default and available since the 10.x code. The router does a route table lookup for the first packet toward a destination and caches it so that it doesn't have to do a route table lookup on each and every packet. If a router actually performs a route table lookup on each and every packet, you can imagine the overhead. This is called process switching and is used when you perform such tasks as debug commands. CEF is a switching type whereby even the first packet gets cached because the switching is performed in hardware.

CEF switching (ip route-cache cef) is now the default and has been available since the 11.x code. In higher-end models, such as the Cisco 12000 GSR routers and Catalyst 6500 switches with MSFC-2 cards, CEF is the default switching type. In the lower-end routers, CEF is an optional switching type and is done in software rather than hardware. You can enable CEF globally and then turn it off on any interfaces that are running features that may interfere with CEF with the following commands:

							ip cef
							interface e0
							no ip route-cache cef
						

CEF uses a Forwarding Information Base (FIB) to make longest match destination-based switching decisions. Think of this as somewhat like the routing table for switching decisions. Each FIB entry points to its Layer 2 rewrite information in the adjacency table. You can read the FIB with the show mls entry cef command, view the contents of the adjacency table with show mls entry cef adjacency, and clear the adjacency table with clear adjacency. There is nothing to turn CEF on or off in its hardware-based form. However, no ip cef disables CEF switching globally in software. Use show ip cef ipaddr and show adjacency [detail] for troubleshooting. To see which packets were dropped or not forwarded by CEF, issue the show cef [dropped|not-cef-switched] command. Distributed CEF (dCEF) synchronizes the line cards to the adjacency table on the route processor; therefore, clear adjacency clears all. If you need to just clear the CEF information on a line card, use clear cef linecard slot# ? instead.

Multilayer Switching (MLS)

MLS is a book in itself, as previously mentioned, but Cisco provides this route switch technology on such platforms as their Cat5000 and 6000 (and now 4000). The 5000 uses the NFFC and the 6000 uses the MSFC along with a PFC.

MLS is a caching technique where the feature card remembers actions taken by the router to shortcut the router the next time. MLS does not take a Layer 2 device and turn it into a router; but it is an advanced form of switching that caches the Layer 3 information.

Multilayer Switching Protocol (MLSP) packets are hello packets sent out by the router. On a Cat5000, for example, this is how the NFFC learns about MLS-capable routers and their MAC addresses. The NFFC identifies candidate packets. It is able to do this based on pattern-matching routines as it looks for packets destined to the MAC addresses gleaned from the hellos. The NFFC identifies enable packets so that it has all the information necessary to rewrite the Layer 2 header as the router did for the first packet. The NFFC shortcuts future packets by rewriting the header itself instead of forwarding it to the router. It also has to decrement and test time to live (TTL) and recalculate the IP header checksum. Basically, the first packet in a flow is sent to the router/Layer 3 engine for a routing decision. If the frame is returned to the switch for forwarding (the destination is reached through another of the switch ports), the switch finishes creation of the cache entry and all other frames are forwarded by the switch without having to go to the Layer 3 engine.

MLS relies on hardware caching to basically shortcut routing, whereas switching routers such as the Cat8500 rely on hardware to perform router functions. A Reduced Instruction Set Computer (RISC)–based CPU handles routing protocols, and intelligent line cards do CEF table lookup and forwarding functions. VLAN features are not directly supported on the 8500s. This is where the ever so popular 6000 series comes in to play.

This chapter's scenario exposed you to VLANs and other related features in a step-by-step practical approach. Make sure you save all configurations and repeat any steps on which you need more practice. As you proved with the lab work, routers bring flexibility and scalability to VLANs.