Ethernet Hubs and Faster Ethernet
Let's Generate Some Excitement
I've already mentioned the challenge to Ethernet's original bus topology that arose when the IBM Token Ring was introduced in 1984. It really is hard for people today to imagine the excitement at the time, when networking was not ubiquitous. I'll reminisce a bit about the days when people could get really excited about a new networking technology. Ah, those were the days!
I was working at Bell Laboratories at the time, and, being a large IBM customer, Bell Labs was invited to various IBM product announcements. Since my specialty was network modeling, I was among the people who attended the Token Ring product announcement at a local IBM branch. Though some Bell Labs installations (or departments) were using Ethernet, the installation that I was working at was an "IBM shop", which means that until very recently - at that time - we worked on an IBM mainframe, using IBM 3270 technology, and when our 3270 terminals were replaced with IBM PCs, the 3270 cabling was "reused" in an ARCNET LAN (which I mentioned in the introduction to this portal).
When the IBM Token Ring was introduced, we were recabled, we got new LAN cards, and we were very excited. This was state-of-the-art technology and we were on the leading edge! Even the new cable connectors were so cool (even if bulky and hard to click), and for those of us who were familiar with the Ethernet AUI connector that was used with 10Base5 (thicknet) Ethernet, whose little clamps that were supposed to fit over the screws were always bending out of shape and getting stuck, the IBM "hermaphrodite" clip was nothing short of brilliant.
But why all this rapture about a technology that is so passe? Especially when this is supposed to be a tutorial about Ethernet and I'm supposed to be talking about Ethernet hubs?! Because one of the real important selling points of the token ring, aside from the touted benefits of the access method itself (which will be discussed in its proper place in the token ring portion of this portal), which did not have collisions, was the star topology, using wiring hubs (Figure 5, though the original IBM Token Ring hub had eight ports). And the wiring was twisted pair (both shielded (STP) and unshielded (UTP)), which was deemed to have benefits over the coax cable used by Ethernet, though the shielded twisted pair was probably just as unwieldy to work with. (On the other hand, the UTP, aside from being cheap and familiar, had simple connectors, like telephone connectors.)
Figure 5: A LAN Hub
Though the token ring is electrically (also referred to as "logically") a ring topology, physically, a ring could be built using a hub, creating a star topology. Every cable that was attached to the hub, to which a computer was connected, actually consisted of both send and receive wires, so that the electrical path made a complete circuit (the hub itself completing the ring), as shown in Figure 6.
Figure 6: A Logical Ring
Further, this new star topology was so convenient for wiring. Via cabling that was hidden in the building walls, all cables could lead to a wiring closet (Figure 7), in which it would be easy to connect or disconnect computers (disconnecting a computer caused the hub to close the ring on the port where the computer left the ring, and you could even hear the electromechanical switch that would either open or close the port). Troubleshooting was also thus facilitated, as was expanding the network. You could easily add more hubs as needed, "daisy-chaining" them to the existing hubs to widen the ring.
Figure 7: Placing Hubs in a Wiring Closet
This topology (and UTP cabling) was something to which Ethernet had to have a response if it were not to be made obsolete by this new, snazzier technology. And that is why I wrote all this about the introduction of the IBM Token Ring. It provides a backdrop to the dramatic impact on networking of the new (at that time) 10BaseT standard for Ethernet, which offered hub-based Ethernet - what I like to call a "bus-in-a-box" - as well as UTP (Unshielded Twisted Pair) cabling.
The Bus-in-a-Box
The 10BaseT (T, for Twisted Pair) version of 802.3, standardized in 1991, was the answer to the star topology offered by the IBM Token Ring. I like to call this topology the bus-in- a-box because, electrically, a hub-based Ethernet behaves exactly like a bus-topology Ethernet. Thus, all frames are received by all connected stations, everyone shares the bandwidth, any connected station can collide with any other connected station, and all collisions can be detected by any station that is connected to the hub. The only practical difference between the bus-in-a-box and a regular bus is the convenience of hub-based wiring and the fact that the stations are connected using UTP wiring.
The bus-in-a-box, or star topology Ethernet, is able to behave just like the bus topology Ethernet because the hub is actually a multi-port repeater. That means that each port on the hub is a repeater, which repeats the signal that is present on the internal bus (inside the box), including enforcing collisions (ensuring that collision information is also repeated). Just as a bus topology Ethernet consists of a single collision domain, even if repeaters are used, the bus-in-a-box, or mutli-port repeater topology Ethernet also consists of a single collision domain.
You can try the interactive animation of an Ethernet hub with the Ethernet Applet.
Beyond 10BaseT
Following the success of the 10BaseT Ethernet, the IEEE continued to develop the 802.3 standards, providing for speeds up to 10 Gbps (and who knows where it will go). The standard was also expanded to include fiber optic cabling in addition to the more commonly found twisted pair. The fiber optic cabling, in addition to offering greater distance between stations and the hubs in wiring closets (and between hubs that are part of the same collision domain), offers greater security and less interference.
The 802.3 standards also include access network Ethernet, also known as Ethernet in the first mile (or last mile). This latter Ethernet is intended as a solution to connect homes and businesses to telecommunications networks and the Internet, like ADSL does, with greater speeds and other benefits. This portal will not address the "last mile Ethernet" technology - it is not really a LAN in the usual sense, but rather a different form of network - specifically an "access network" that utilizes some of the characteristics of Ethernet. All of the flavors of IEEE 802.3 are detailed in the IEEE Std 802.3-2008 documents, which, as of this writing, are the latest version of these IEEE standards. If you look at Section I, you will see that what is common to all types of 802.3 is the frame format.
I won't expand upon the dry details of the differences in the different physical characteristics of the different speed (and media) versions of Ethernet, but we will look at another development in Ethernet, and that is Ethernet switches. Just as the hub architecture revolutionized the bus-based topology Ethernet, the switch architecture rescued Ethernet from the perils of the single collision domain.
One of the disadvantages of the hub-based Ethernet (or of the original bus-based topology) is the fact that the Ethernet consists of one collision domain. If there are too many very active stations on the Ethernet, the number of collisions becomes unweildy, and transmission can be delayed, resulting in unacceptable performance. This aspect of Ethernet is a result of its access method, which is one of the points of "attack" when token ring was first introduced. The deterministic nature of the token ring access method was considered a great advantage over Ethernet.
When Ethernet was first developed, the types of data that were transmitted were basically text and text. The message sizes were tiny in comparison to the message sizes of today, where a message might include a video with a soundtrack. If the advantage of the deterministic access method of token ring was pretty much theoretical in the 1980s, it was very real as the content sent over the networks became more sophisticated. In the chapter on Ethernet switches we shall see how Ethernet switching has virtually eliminated the problems caused by collisions, but before we get to switches, I must explain Ethernet addresses, which we'll look at in the following chapter about the Ethernet frame format.
