6.3 External Routers
6.3.3 Trunk-connected routers
As technologies such as ISL became more common, network designers began to use trunk links to connect routers to a campus backbone. Figure illustrates an example of this approach.

Although any trunking technology such as ISL, 802.1Q, 802.10, or LAN Emulation (LANE) can be used, Ethernet-based approaches are most common (ISL and 802.1Q). Figure uses ISL running over Fast Ethernet. The solid lines refer to the single physical link running between the Catalyst Switch and the router. The dashed lines refer to the multiple logical links running over this physical link using subinterfaces.

The primary advantage of using a trunk link is a reduction in router and switch ports. Not only can this save money, it can also reduce configuration complexity. Consequently, the trunk-connected router approach can scale to a much larger number of VLANs than a one-link-per-VLAN design.

However, the trunk-connected router configuration has disadvantages, including the following:

  • There is a possibility of inadequate bandwidth for each VLAN.
  • Additional overhead on the router can occur.
  • Older versions of the Cisco IOS Software support only a limited set of features on ISL interfaces.

With regard to inadequate bandwidth for each VLAN, consider, for example, the use of a Fast Ethernet link where all VLANs must share 100 Mbps of bandwidth. A single VLAN could easily consume the entire capacity of the router or the link (especially in the event of a broadcast storm or Spanning-Tree problem).

With regard to the additional overhead on the router caused by using a trunk-connected router, not only must the router perform normal routing and data-forwarding duties, it must also handle the additional encapsulation used by the trunking protocol. Consider ISL running on a high-end router as an example. These software-based routers have many different switching modes, a term that Cisco uses to refer generically to the process of data forwarding in a router. Don't confuse the term switching here with how it is normally used throughout this course. These software-based routers use the term switching to refer to the process of forwarding frames through the box, regardless of whether the frames are routed or bridged.

Every Cisco router supports multiple forwarding techniques. Although a full discussion of these is not appropriate here, an analogy can be made to make the point: think of switching modes as gears in an automobile transmission. For example, just as every car has a first gear, every Cisco router (including low-end routers) supports a feature called process switching. Process switching relies on the CPU to perform brute-force routing on every packet. Just as first gear is useful in all situations (uphill, flat roads, rain, snow, dry, and so on), process switching can route all packets and protocols. However, just as first gear is the slowest in a car, process switching is the slowest forwarding technique for a router.

Every router also has a second gear-this is referred to as fast switching. Taking advantage of software-based caching techniques provides faster data forwarding. However, just as second gear is not useful in all situations (going up a steep hill, starting at a traffic stop, and so on), fast switching cannot handle all types of traffic (for example, many types of Systems Network Architecture [SNA] traffic).

Finally, just as high-end automobiles offer fancy five-speed transmissions, high-end routers offer a variety of other switching modes. These switching modes are known by names such as autonomous switching, silicon switching, optimum switching, and distributed switching. Think of these as gears three, four, and five in a transmission-they can allow you to move very quickly, but can be useful only in ideal conditions and very limited situations (that is, dry pavement, a long country road, and no police!).

Getting back to the example of an ISL interface on a high-end router, these routers normally use techniques such as optimum switching and distributed switching to achieve data-forwarding rates from 300,000 to over 1,000,000 packets per second (pps).

Several performance figures are included in this chapter to allow you to develop a general sense of the throughput you can expect from the various Layer 3 switching options. Keep in mind that throughput numbers are dependent on many factors such as configuration options, software version, and hardware revision.

When running ISL on an interface, that interface becomes limited to second gear (fast switching). Because of this restriction, ISL routing is limited to approximately 50,000 to 100,000 pps on a high-end router (and considerably less on many lower-end platforms).

The third disadvantage of the trunk-connected router design is that older versions of the Cisco IOS Software support only a limited set of features on ISL interfaces. Although most limitations were removed in 11.3 and some later 11.2 images, networks using older images need to carefully plan the inter-VLAN routing in their network. Some of the more significant limitations prior to 11.3 include the following:

  • Earlier versions support only IP and IPX. All other protocols (including AppleTalk and DECnet) must be bridged. Inter-VLAN bridging is almost always a bad idea because IPX supports only the novell_ether encapsulation (Novell refers to this as Ethernet_802.3).
  • Hot Standby Router Protocol (HSRP) is not supported. This can make it very difficult or impossible to provide default gateway redundancy.
  • Secondary IP addresses are not supported.

The example in Figure configures a Fast Ethernet port to perform ISL Routing for three VLANs.