WANs are collections of interconnected local area network links of which the smaller ones are called APs, and the larger ones are called HWANs. These links go through routers that were installed as part of the copper network in the 1950s (prior to that, these routers were installed in the earliest “radio handsets”), and each one can have up to 4 hops of internal and external link density.
The problem is that most of the APs in today’s Ethernet networks are concentrated in close proximity to one another; they are a distributed function of a single AP design, making them very sensitive to variation in wire density. So all the local links must be fast enough, but as we’ll see later, the networks they serve in turn are designed with slow-release connectors, which means there is little buffer time to re-configure the local links as new links are brought up, while also using resources for this purpose like SD-WAN services which you can find explained at sites such as https://www.fortinet.com/resources/cyberglossary/sd-wan-explained. So this means that any time a local link takes a bottleneck – either due to a hardware fault or its bandwidth being restricted – it must get re-allocated to the fastest link in the network, and that results in traffic congestion for all the other links that can use it.
I wrote about this in more detail in “How Network Design Differs from Physics” in April 2008, when I outlined the more subtle but real connection between Ethernet and PHY design.
At this point, we’ve created a two-way tunnel that goes from one router to another, but we need to build out to three or more links so that it can span multiple communities. We do that with the use of loopback (LR) routers, which are a key concept in modern Ethernet networking.
Back in the day, there were several sorts of loops in a local area network, including those where all the links in the network sat in straight lines with few edges (i.e. single-end-address) and a path count that was low. However, since our local links were distributed, it was more common for a link to be found at a split point between two communities: the inside of one community was called the “inside” of the network, and the outside of the other was called the “outside”.
With the arrival of Ethernet, there was an expectation that all loops would be built around these splits, creating large networks with multiple links with low paths. This worked well for a couple of decades, but as we’re about to see, many systems didn’t have sufficient bandwidth to support the network’s growth and use. With the advent of DOCSIS-2 and 10BaseT (not to be confused with the 40-bit blocks that are used as a prefix in 10BaseT, although 10BaseT could become 10BaseT), more developers began finding that they needed more links to handle increasing network traffic volumes.
As a result, loopback (LR) routers were introduced, essentially removing the split point from the routing diagram and allowing the networks to stay as two-way while adding more links.
Now we’re going to see a typical network with several communities, all managed by a separate router that is responsible for keeping track of the topology of all the communities. The paths between communities are represented by arrows in the network diagram, and all nodes in the network can access any of the communities with the help of data fields known as “pipes”. The networks themselves are really just high-capacity data transfers between communities.