Question about the way networks are described vs their modern implementations

Question

I am a networking student and I am currently pursuing my CCNA. I am primarily enrolled in the Cisco NetAcad.

My question is in regards to how certain networks are described in these courses compared to there modern implementations.

Ethernet has always confused me, it is always described as a shared medium, but in a modern network almost all wired connections are technically point to point (host to switch, router to switch, switch to switch) on the physical layer. From my understanding the original ethernet was a bus topology and all of the devices truly shared the same medium, is this where this way of describing it comes from?

Also the way Cisco described point to point connections is also quite confusing. I know the most basic definition is a connection between 2 devices only, but cisco describes many different versions of this including serial p2p and p2poe. I think I understand this but the way they are explaining this is muddying the waters for me so to speak.

If anyone could help me understand these concepts better that would be great!

Answer 1

A lot of educational material does a poor job of distinguishing between original ethernet and modern ethernet.

Originally Ethernet 10BASE-2 and 10BASE-5 used a shared coaxial cable. The reach of the system and number of devices could be increased by deploying repeaters. Repeaters were pretty dumb devices, they simply detected which side was transmitting and amplified the signal. So on an ethernet network with repeaters it was still only possible for one device to succesfully transmit at a given time.

Then 10BASE-T came along, it has dedicated pairs for each direction, so in principle it could support full duplex. However in practice 10BASE-T was nearly always used in a half-duplex mode for two reasons.

1. There was no way to indicate to the end device that full duplex mode was in use, and a mismatched duplex configuration could cause big problems.
2. 10BASE-T was nearly always used with multiport repeaters known as hubs. These hubs were again dumb devices, so they could only support a single device transmitting at any given time.

In addition to repeater hubs, there also existed "bridges", these operated at a higher level, they could receive frames, filter them by destination mac address and queue them for further transmission. Initially bridges were implemented in software using general purpose computers, but around the same time that 10BASE-T appeared on the scene Kalpana introduced the "etherswitch", a fast hardware bridge. Initially, switches were expensive but over time they came down in price.

Then fast Ethernet came along, this changed a couple of things

1. There was now an auto-negotiation protocol. So switches and end devices could negotiate full duplex mode without error-prone manual configuration.
2. Connecting fast Ethernet to regular Ethernet required a bridge, a dumb repeater hub could only connect devices of the same speed.

In the early days of fast Ethernet there did exist some "dual-speed hubs", which essentially behaved like a 10 megabit hub and a 100 megabit hub connected by a two port bridge but over time, as hardware costs came down, these gave way to full 10/100 switches.

By the time gigabit Ethernet came along, hubs were pretty much obsolete. The gigabit Ethernet specification does contain support for hubs and a half-duplex mode for them but my understanding is it was included more for political than technical reasons (there were arguments about whether a standard that did not support CSMA/CD could still be considered part of Ethernet). If a gigabit hub was ever sold I can't find any evidence of it.

10 gigabit and newer Ethernet standards (including 2.5G and 5G) officially dropped support for hubs and half-duplex mode altogether.

The result of this is that while most Ethernet hardware generally does still support CSMA/CD, it rarely uses it.

Despite the lack of CSMA/CD, a full duplex Ethernet connection is still treated a multipoint network by the IP stack on a host or router. Yes it could be a direct link to another host or router router but it could also be connected to an Ethernet switch or switches with multiple other hosts or routers.

Answer 2

(Ethernet) is always described as a shared medium

A shared electric (coax) medium on the physical layer is long obsolete. It vanished when 10BASE-T took over from 10BASE2 in the early 1990s. Some might argue that repeated 10BASE-T or 100BASE-TX (via hubs) is also shared logically but that vanished in the late 1990s as well.

Ethernet still looks shared on the data link layer but it isn't really. Ignoring the less common passive optical variants, all PHYs are point to point today. (There has been some recent effort to create specialized shared, single-pair variants though - not sure if that is still under way.)

the original ethernet was a bus topology and all of the devices truly shared the same medium, is this where this way of describing it comes from?

Most likely, yes.

Sadly, educational material seems to get updated very rarely, so students are still taught half-duplex Ethernet, classful networking and such from decades past - but notable for historical reasons only.

cisco describes many different versions of this including serial p2p and p2poe

Not sure what the context is but Cisco traditionally supports high-speed serial links, not related to Ethernet. You might also be referring to PPPoE or PPTP which create P2P links over Ethernet or over TCP/GRE respectively - they are no Ethernet variants either.

For a beginner it's important to grasp the basic, common concepts first. If you try to look at all the stuff that is around it's very confusing for sure.