This chapter demonstrated both the basic and advanced spanning-tree configurations. The basic configuration process can be summarized as follows:
You should place your root bridge so that the calculated STP topology matches the major traffic paths of your network.
If you have multiple VLANs, a spanning-tree instance exists per VLAN, and you should adjust spanning-tree parameters to ensure any redundant links are load shared across. The following mechanisms exist for configuring spanning tree load sharing:
Modifying per VLAN costs is the recommended method of load sharing because root bridge placement affects load sharing only for switches connected directly to the root and port ID is seldom the deciding factor for path selection. Port ID is used only when two switches are connected by multiple connections. In this scenario you should consider combining the physical connections into an EtherChannel bundle, which is represented as a single logical Layer 2 connection to STP. EtherChannel handles failure and load sharing much more efficiently than STP.
Spanning tree does have considerable convergence delays, with the default convergence times ranging from 30 seconds (direct failure) to 50 seconds (indirect failure). Spanning-tree timers are calculated using a number of parameters, including the configurable Hello timer and network diameter parameters. Reducing your Hello timer speeds up how quickly the network learns about a failure; however, you must consider the effect on CPU load of the switch before reducing the Hello timer. By default, spanning-tree timers are configured to accommodate large Layer 2 networks with a network diameter of seven bridges or less. If your network diameter is smaller, you can reduce your spanning-tree timers to reflect this. Using the set spantree root command with the optional dia and hello parameters is the recommended method of adjusting spanning-tree timers because it uses the IEEE 802.1d calculations for each timer.
Cisco has developed a number of spanning-tree enhancements that are designed to reduce spanning-tree convergence in certain situations. These enhancements include the following: