5.6. Behavior of a Spanning Tree Catenet
Many approaches can be used to maintain proper end station behavior in a bridged catenet. The decision to use a spanning tree arises mostly from the simplicity of the algorithm, which allows low-cost bridges to be easily built. When the catenet consists of high-speed LANs in a local environment, this seems a reasonable decision. In different environments, other factors must be considered. For example, the STP prohibits load-sharing across multiple paths to the same destination. This may be acceptable in a bridged catenet, but is generally intolerable in a complex WAN (such as the Internet), where link bandwidth is expensive and needs to be used to its maximum capabilities.
In this section, we look at the behaviors exhibited by catenets using a spanning tree topology.
5.6.1. Maintaining the Link Invariants
A single, standalone (unbridged) LAN exhibits the following invariant behaviors:
Sequential delivery: It is not possible for frames to be sent by a source on the LAN and received by a destination in a different order from that in which they were sent. It is possible that some frames might not be delivered, but there is no way for the sequence of frames to be changed.
Non-duplication: Similarly, there is no means for a station to send a single frame and have another station receive multiple copies of that frame.
These are so-called hard invariants discussed in Chapter 2. There are no provisions made in the protocols (for example, sequence ...