SONET network is divided into three categories: linear, ring, and mesh networks, as shown in Fig.1.

1. Linear Networks: A linear SONET network can be point-to-point or multipoint.

Point-to-Point Network: A point-to-point network is normally made of an STS multiplexer, an STS demultiplexer, and zero or more regenerators with no add/drop multiplexers, as shown in Fig.2. The signal flow can be unidirectional or bidirectional, although Fig.2. shows only unidirectional for simplicity.

Multipoint Network: A multipoint network uses ADMs to allow communications between several terminals. An ADM removes the signal belonging to the terminal connected to it and adds the signal transmitted from another terminal. Each terminal can send data to one or more downstream terminals. Fig.3 shows a unidirectional scheme in which each terminal can send data only to the downstream terminals, but a multi-point network can be bidirectional, too.

2. Ring Networks: SONET rings can be used in either a unidirectional or a bidirectional configuration. In each case, we can add extra rings to make the network self-healing, capable of self-recovery from line failure.

Unidirectional Path Switching Ring: A unidirectional path switching ring (UPSR) is a unidirectional network with two rings: one ring used as the working ring and the other as the protection ring. The same signal flows through both rings, one clockwise and the other counterclockwise. It is called UPSR because monitoring is done at the path layer. A node receives two copies of the electrical signals at the path layer, compares them, and chooses the one with the better quality.

If part of a ring between two ADMs (Automatic Protection Switching) fails, the other ring still can guarantee the continuation of data flow. UPSR, like the one-plus-one scheme, has fast failure recovery, but it is not efficient because we need to have two rings that do the job of one. Half of the bandwidth is wasted. Fig.4 shows a UPSR network.

Although we have chosen one sender and three receivers in the figure, there can be many other configurations. The sender uses a two-way connection to send data to both rings simultaneously; the receiver uses selecting switches to select the ring with better signal quality. We have used one STS multiplexer and three STS demultiplexers to emphasize that nodes operate on the path layer.

Bidirectional Line Switching Ring: Another alternative in a SONET ring network is a bidirectional line switching ring (BLSR). In this case, communication is bidirectional, which means that we need two rings for working lines. We also need two rings for protection lines. This means BLSR uses four rings.

If a working ring in one direction between two nodes fails, the receiving node can use the reverse ring to inform the upstream node in the failed direction to use the protection ring. The network can recover in several different failure situations that we do not discuss here. Note that the discovery of a failure in BLSR is at the line layer, not the path layer. The ADMs find the failure and inform the adjacent nodes to use the protection rings. Fig.5 shows a BLSR ring.

3. Mesh Networks: One problem with ring networks is the lack of scalability. When the traffic in a ring increases, we need to upgrade not only the lines, but also the ADMs. In this situation, a mesh network with switches would probably give better performance. A switch in a network mesh is called a cross-connect. A cross-connect, like other switches we have seen, has input and output ports.

In an input port, the switch takes an OC-n signal, changes it to an STS-n signal, demultiplexes it into the corresponding STS-1 signals, and sends each STS-1 signal to the appropriate output port.

An output port takes STS-1 signals coming from different input ports, multiplexes them into an STS-n signal, and makes an OC-n signal for transmission. Fig.6 shows a mesh SONET network, and the structure of a switch.