Token ring – IEEE 802.5

A token ring is an outdated form of network tech­nol­o­gy, and today Ethernet is the most prevalent type of wired LAN con­nec­tion. But that doesn't make the old tech­nol­o­gy any less in­ter­est­ing. Older methods, now confined to the realms of history, can also help us un­der­stand modern networks and why tech­nol­o­gy has developed as it has.

However, a token ring is not really a ring network, also referred to as a ring topology, even if the name suggests this. In fact, the only ring created is a logical one, not a physical one. We will explain the dif­fer­ences below.

Even before the in­tro­duc­tion of the token ring, there were, of course, other computer networks. These were not available to the public, like ARPANET, but were mostly developed and used by in­di­vid­ual in­sti­tu­tions for their own purposes and were therefore not stan­dard­ized. In the early 1980s, the companies Procom, Apollo Computer, and Proteon were the first to offer token ring networks on a larger scale. IBM then took the matter up and launched its own product in the middle of the decade.

Due to how wide­spread IBM PCs became, the token ring also began to quickly take over the tech market – and was directly slowed down again. Ethernet tech­nol­o­gy, which was already developed in the 1970s, also gained more and more pop­u­lar­i­ty in the 1980s, partly thanks to the support of Intel and Xerox. Ethernet beat the com­pe­ti­tion between the two tech­nolo­gies by in­tro­duc­ing low-cost, thin co-axial cables. IBM had charged such high fees for the license of the token ring process that no in­ex­pen­sive products could come onto the market.

While IBM initially tried to convince the public of the ad­van­tages of a token ring, they gave up at the end of the 1990s: Although the IEEE had approved models with 100 Mbit/s and 1,000 Mbit/s, the former was marketed only slightly, and with the latter not a single product made it onto the market.

The question of what a token ring network is easier to answer if you know what it is not: a ring topology. A ring topology in the physical sense consists of a circular arrange­ment of computers. Each net par­tic­i­pant is connected to its neighbors to the left and right, so that the network is a closed ring. As soon as a computer within the LAN fails or a con­nec­tion is otherwise dis­con­nect­ed, the whole network breaks down, and the con­nec­tion is lost. If another computer wants to par­tic­i­pate in the network, the LAN must be in­ter­rupt­ed for at least a short moment.

A token ring works somewhat dif­fer­ent­ly to other ring topolo­gies, which is why it’s said that this tech­nol­o­gy is based only logically on a ring topology. The token ring topology uses Mul­ti­sta­tion Access Units (MAUs), which allow a star-shaped con­nec­tion of the con­nec­tions involved. The dis­trib­u­tor is a node that is connected to all computers on the network. There is no direct con­nec­tion between the in­di­vid­ual computers.

Nev­er­the­less there is still a logical ring involved, which is due to the physical star structure, because the data trans­mis­sion takes the form of a ring – on an abstract level. Although the data is re­peat­ed­ly trans­port­ed to the MAU, it is not sent from there to a specified sub­scriber, but simply to the next computer in the fixed sequence.

Issues can occur in any network, and in most there is the risk that some would cause the network to break down. In a token ring, all computers are standby monitors (SM) by default, but you usually check on the network in the role of an active monitor (AM). Each station can take on this role. The decision as to which station will become AM – and so also which ones remain as SM – is made by a fixed procedure: monitor con­tention, also called token claiming. The procedure is initiated when a station has de­ter­mined that the current AM no longer transmits the mandatory active monitor present frame via the ring.

When this happens, the protocol requires a choice for a new active monitor: The first station that has noticed the mal­func­tion of the AM begins to send a so-called claim token frame. The next station in the ring compares the sender's MAC address with its own. If the address of this second network sub­scriber has a higher value, it replaces the frame with its own. At the end only one station remains: If its frame has circled the ring three times without any other station in­ter­ven­ing, a new active monitor is de­ter­mined.

The active monitor is re­spon­si­ble for the smooth running of the token ring, but standby monitors can also intervene in some cases. In the following scenarios, the token ring network can heal itself:

As soon as a token passes the active monitor, it sets a timer. If the time span (10 mil­lisec­onds) has elapsed before the token reaches it again, the AM detects a problem in the network and generates a new, free token.

If one station sends a packet and the recipient station would fail at this very moment, the packet would circle endlessly on the ring, since no station iden­ti­fies itself as the recipient. When the packet passes the AM for the first time, it converts a specific bit. If the packet passes the AM again, he rec­og­nizes from the bit that the packet has not reached his recipient. Only the AM can set the bit to 1 and de­ter­mines that it must already have come into contact with the packet. The active monitor destroys the package and creates a new, free token.

In a normal ring topology, the failure of a station or con­nec­tion means the complete failure of the network. With a token ring topology, however, the MAU can simply bridge the interface. Faulty stations are detected by a direct neighbor: To do this, the par­tic­i­pant located directly behind the damaged computer sends test frames to the neighbor. All other stations enter a waiting mode: no station sends data. If the faulty node detects that its neighbor has submitted a complaint, it starts an error analysis and takes itself off the network.

However, the error can also be at the original station: The fact that the computer can no longer access data may also be due to a faulty network card. So, the par­tic­i­pant sets a timer that would give the neighbor enough time to perform a self-test. If no frame arrives after the ex­pi­ra­tion of the timer, the station assumes an error and starts a test.

The faulty station – if it was not the AM – is simply bypassed during trans­mis­sion until a solution has been found. If it is the AM, the token claiming starts. When every­thing works again, the AM creates a new token and mains operation can continue normally.

The ad­van­tages and dis­ad­van­tages of a token ring can be seen in a com­par­i­son with Ethernet systems. The fact that col­li­sions cannot occur makes the token ring an in­ter­est­ing concept: In a (half-duplex) Ethernet network, col­li­sions caused by several stations trans­mit­ting at once are likely. With the help of CSMA/CD you can control how these col­li­sions are handled, but they reduce the speed anyway. In a token ring, however, col­li­sions cannot occur. The token-passing procedure prevents several stations from trans­mit­ting si­mul­ta­ne­ous­ly.

This com­pen­sates for the token ring’s lack of speed. While token ring tech­nol­o­gy was stopped in its de­vel­op­ment at 16 Mbit/s, Ethernet networks with 100 Mbit/s already existed at that time. If you compare a simple ring topology with the token ring via MAUs, the latter has the advantage of flexible nodes. Adding or removing in­di­vid­ual computers from the network is no problem. Even if in­di­vid­ual stations fail un­ex­pect­ed­ly, the network is not en­dan­gered. This in com­bi­na­tion with effective trou­bleshoot­ing makes a token ring a very stable system.