A network can be a simple and direct as a connection between just two points (nodes), or it can incorporate tens, hundreds, thousands and even more nodes. How these nodes are connected, how the network functions under ideal and stressed situations and traffic levels, and the key attributes of each network configuration is largely defined by the network topology. Part 1 of this FAQ looks at basic, widely used point-to-point topologies as well as more-complex ones.
Q: Why is a network carrying information in digital format really an analog topic?
A: The unavoidable reality of all networks is that for any node to connect to another node requires analog circuitry such as line drivers and receivers, or RF transmitters and receivers. These must function in the real, analog-based world. In wired (copper and optical) systems, these drivers and receivers must present the correct voltage/current to the network and receive (and tolerate) signals with wide-ranging voltage, current, and power (despite the nominal setting). For wireless networks, there are issues of transmitter output power, and received signal strength. Further, all networks deal timing issues such as jitter as well as noise. All of these are analog topics, regardless of the fact that the signals are representing information in digital format.
In formal terms, regardless of the specifics of the implementation of the seven-layer Open System Interconnection (OSI) Model developed by the International Organization for Standardization (ISO), Figure 1, there is always Layer 1, the Physical Layer. This is where actual connections to the channel medium (wired and wireless) are established and maintained. While Layers 2 through 7 are digital, Layer 1 is the real, analog world, with all the issues that the analog world brings.
Q: How does wired versus wireless affect network topologies?
A: Topologies can be implemented either using hard-wired interconnections (copper or optical fiber) or wireless ones. The tradeoffs in cost, speed, reliability, distance, BER and other performance factors differs with the wired versus wireless physical realization.
Q: What is the simplest network?
A: It’s obvious: the point-to-point, two-node network is the simplest, as it connects points A and B and that’s all it does, Figure 2. Although this may seem like a limited network in terms of performance, it is all that is needed for a basic interconnect function, such as connecting a probe to a test instrument. The legendary RS-232 interface is a basic point-to-point network.
What are the possibilities beyond point-to-point in complexity and performance?
A: There are several choices: star, bus, ring, and mesh, as well as combinations of these. Again, these topologies can be wired or wireless. What changes are the performance capabilities, difficulties of installation and debug, complexity of the needed circuitry, and challenges of network management. As nodes are added to any network, management of the nodes and their interaction becomes an issue.
Q: What is the simplest option after a point-to-point interconnection?
A: The next logical step is the bus, a common pathway connecting multiple nodes, where each node has a unique address, Figure 3. This is also referred to as a multidrop configuration. In some bus designs, all nodes are “equal”; in others, one node is a master and all other are slaves. The bus itself is sometimes called a backbone.
Q: What are the attributes of the bus topology?
A; First, it’s relatively straightforward to add additional nodes. It’s also easy for a master node (if used) to broadcast a message to all or a select subset of the other nodes. But getting messages from an equal-rated or slave node to other nodes or the master node is more difficult. Among the approaches, the master can sequentially poll each node in turn, asking if there is a message (query/response). Or, each node can sequentially have a time slot in which to “speak up” and say if it has/doesn’t have any new message, which requires high-accuracy node timing.
Either way, there will be a lot of traffic on the bus that is related to management rather than actual new data, so it consumes considerable bus bandwidth with overhead. However, both approaches are deterministic, though, in the sense that you can calculate the worst-case delay for a new message from one node to another, which is important in many situations.
If a hard-wired bus, there must be some way to electrically disconnect each node from the bus when that node is in a quiescent mode, typically done by using three-state drivers; this is what is done with RS-285, a widely used multidrop-bus standard.
Q: What’s an alternative to those bus-based query/response or time slot-based management?
A: The most-common alternative is based on the concept of collision avoidance. In principle, each node can begin to transmit when it has new data, but this would result in collisions when two nodes attempt to do so at the same instant. The solution is to have a node first check if any other node is active on the bus. If so, the node waits (backs off) a random amount of time, then retries again.
Formally, this technique is known as CSMA/CD (Carrier Sense Multiple Access/Collision Detection) and used as the basis for Ethernet. It is efficient in terms of use of bus bandwidth. However, the amount of time that a node may have to wait to get its “turn” is not deterministic but instead is random, and that can be a drawback in some situations, especially if the overall bus speed is low, the data load is high, or the nodes have long messages. To overcome the problem of long messages, each node is only allowed to send to send a packet up to a certain maximum length, then must relinquish its turn so as to allow others to use the bus.
Part 2 of this FAQ will look at other widely-used topologies: ring, star, mesh, and tree.
References
Lucid Software, Inc., “What is a Network Diagram”
Computer Hope, “Computer terms, dictionary, and glossary”
