Token Ring

IBM Token Ring
Company type	Proprietary technology
Industry	Computer networking
Founded	1984; 42 years ago
Founder	IBM
Headquarters	United States
Products	Token Ring LAN technology

Token Ring is a physical and data link layer computer network technology designed for constructing local area networks. Developed by IBM in 1984 and officially standardized as IEEE 802.5 in 1989, it operates by circulating a distinctive three-byte token frame around a logical ring of computers or servers. This token passing mechanism guarantees equitable network access for all devices, preventing collisions that are typical in contention-based protocols.

While Token Ring found considerable use within business networks, it was eventually supplanted by advancing Ethernet systems. The final official Token Ring standard, known as Gigabit Token Ring (IEEE 802.5z), was released on May 4, 2001.

History

During the early 1970s, several local area network solutions appeared, including the Cambridge Ring, which showcased the effectiveness of a token passing ring topology. At the IBM Zurich Research Laboratory, engineers Werner Bux and Hans Müller spearheaded the creation of IBM's Token Ring technology. Simultaneously, pioneering developments at MIT contributed to the Proteon ProNet-10 Token Ring network introduced in 1981. That year also saw workstation manufacturer Apollo Computer launch their proprietary Apollo Token Ring (ATR) system using 75-ohm RG-6U coaxial cabling. Proteon later advanced their design to operate over unshielded twisted pair cables.

1985 IBM launch

On October 15, 1985, IBM unveiled its proprietary Token Ring solution. Operating at , it supported connectivity across IBM PCs, midrange systems, and mainframes. The architecture utilized a star-wired physical layout with shielded twisted-pair cabling and subsequently became the foundation for the IEEE 802.5 standard.

IBM marketed Token Ring LANs as outperforming Ethernet, especially under conditions of heavy traffic; however, these assertions were debated within industry circles.

In 1988, the IEEE 802.5 working group ratified a faster Token Ring standard running at . Subsequent attempts to increase speeds to were introduced but achieved limited market penetration. A standard for was approved in 2001; nonetheless, no commercial implementations followed as Fast Ethernet and Gigabit Ethernet technologies dominated local networking.

Comparison with Ethernet

Both early Ethernet and Token Ring employed shared communication media but utilized distinct channel access methods. Advances in Ethernet, such as switching and point-to-point links operating in full-duplex mode, have largely diminished these differences.

Notable contrasts between Token Ring and traditional Ethernet include:

Token Ring employs a more deterministic access approach compared to Ethernet’s contention-driven CSMA/CD protocol.
Unlike Token Ring—which necessitates additional hardware and software for direct cable connections—Ethernet supports direct NIC-to-NIC links using crossover cables or Auto MDI-X functionality.
Collisions are avoided in Token Ring by using a single circulating token and early token release; legacy Ethernet relies on carrier-sense multiple access with collision detection, where hubs can still cause collisions.
Token Ring network interface cards (NICs) incorporate intelligence for speed auto-negotiation and routing, enabling them to function with numerous passive Multistation Access Units (MAUs). Ethernet NICs have limited compatibility with passive hubs and continue to experience collision issues.
The access priority feature in Token Ring allows specific nodes preferential token access; traditional unswitched Ethernet treats all devices equally.
Token Ring supports multiple identical MAC addresses—a capability utilized by S/390 mainframes—whereas switched Ethernet does not permit duplicate MAC addresses.
Interfaces for Token Ring required specialized processors and licensed MAC/LLC firmware, resulting in costs approximately triple those of comparable Ethernet adapters based on Intel 82586 chips.
Although both technologies initially used expensive cabling, Ethernet’s adoption of unshielded twisted pair standards (10BASE-T and 100BASE-TX) provided a cost benefit that boosted its market growth.
Overall expenses favored Ethernet due to lower prices for router interfaces and network cards; the widespread adoption of Ethernet switches further reinforced this cost advantage.

Operation

In a Token Ring LAN, stations are logically arranged following a ring topology, where data flows in sequence from one device to the next through a circulating control token. Comparable token passing systems are implemented by technologies such as ARCNET, token bus, 100VG-AnyLAN (802.12), and FDDI, which theoretically offer benefits compared to the CSMA/CD protocol used in early Ethernet networks.

Access control

The mechanism for data transmission works as described below:

Empty frames perpetually circulate around the ring.
When a device intends to send data, it captures the token and transmits its frame.
Each subsequent station inspects the frame; the recipient copies the data and resets the token to zero.
Once the frame returns, the sender detects the reset token, confirming successful delivery, and removes the message.
The frame then continues its journey as an empty frame, ready for future transmissions.

Multistation Access Units and Controlled Access Units

From a physical standpoint, Token Ring networks are wired in a star configuration with MAUs positioned centrally to interconnect each station in a loop.

MAUs operate similarly to hubs or switches; since Token Ring avoids collisions, most MAUs functioned primarily as hubs. Although Token Ring uses LLC, it supports source routing beyond local segments. Most MAUs defaulted to 'concentration' mode, but advanced models like IBM's 8226 could serve as splitters rather than just concentrators.

IBM later developed Controlled Access Units (CAUs), which accommodated multiple MAU modules known as Lobe Attachment Modules (LAMs). CAUs provided Dual-Ring Redundancy for alternative routing in case of port failures, modular concentration through LAMs, and support for multiple interfaces. This innovation improved network reliability and enabled remote management capabilities beyond those of unmanaged MAU hubs.

Cabling and interfaces

The typical cabling employed was IBM "Type-1," a robust two-pair 150 ohm shielded twisted pair cable forming the foundation of IBM's structured cabling system aimed at widespread deployment. The connectors used were unique hermaphroditic types called IBM Data Connectors, also informally known as Boy George connectors. These connectors are relatively bulky—requiring at least one rack panel space—and somewhat delicate, but they provide genderless connections and superior shielding compared to standard unshielded 8P8C connectors. Computer interfaces commonly utilized female DE-9 connectors. Additionally, type 2 and type 3 cable variants were available.

In later Token Ring versions, support expanded to include Cat 4 cabling with 8P8C (RJ45) connectors on MAUs, CAUs, and network interface cards (NICs). Many NICs offered compatibility with both 8P8C and DE-9 connectors to maintain backward compatibility.

Token

When the network is idle and no device is transmitting, a distinctive three-octet token frame continuously circulates around the ring, passed sequentially from one station to the next until a device requires transmission access. This token frame comprises a start delimiter, an access control byte, and an end delimiter.

Start Delimiter	Access Control	End Delimiter
8 bits	8 bits	8 bits

Abort frame

This frame type is employed by a sender to terminate an ongoing transmission prematurely.

SD	ED
8 bits	8 bits

Data

Data frames are used to transport upper-layer protocol payloads, whereas command frames carry control information without any payload. The total frame length depends on the size of the Information field.

SD	AC	FC	DA	SA	PDU from LLC (IEEE 802.2)	CRC	ED	FS
8 bits	8 bits	8 bits	48 bits	48 bits	Up to 4500 × 8 bits	32 bits	8 bits	8 bits

Start delimiter: A unique bit sequence indicating the beginning of a frame, characterized by special code violations (J and K) in Differential Manchester encoding.

J	K	0	J	K	0	0	0
1 bit	1 bit	1 bit	1 bit	1 bit	1 bit	1 bit	1 bit

Access control: This byte contains priority bits (P), token indication (T), monitor flag (M), and reservation status (R) which regulate frame priority and handling.

+	Priority (bits 0–2)				Token (bit 3)	Monitor & Reservation (bits 4–7)

Frame control: A single byte that specifies the frame category—either data or control—and additional control information.

+	Frame type (bits 0–1)				Control Bits (bits 2–7)

Frame type – 01 represents an LLC frame compliant with IEEE 802.2 (data); 00 denotes a MAC control frame without data.

Destination address: Six-byte physical address identifying the intended recipient(s).
Source address: Six-byte physical address of the sender’s network interface.
Data field: Variable length, up to 4500 bytes, containing MAC management or upper-layer protocol data.
Frame check sequence: A 32-bit CRC used for error detection.
End delimiter: Marks the conclusion of the frame with a specific bit pattern including error and intermediate flags.

J	K	1	J	K	1	I	E
1 bit	1 bit	1 bit	1 bit	1 bit	1 bit	1 bit	1 bit

Frame status: An 8-bit field indicating whether the receiver has acknowledged and copied the frame.

A	C	0	0	A	C	0	0

A = Address recognized

C = Frame copied

Active and standby monitors

Within a Token Ring network, each station operates either as an active monitor (AM) or as a standby monitor (SM). Only one active monitor exists per ring, chosen through an election mechanism known as monitor contention.

Monitor contention is initiated under circumstances such as:

loss of ring signal,
absence of an active monitor,
expiration of a timer (for example, when no token frame is detected within 7 seconds).

Upon such events, a station transmits a claim token frame to assert itself as the new active monitor. If this token completes a full circulation and returns to its originator, that station assumes the active monitor role. In cases where multiple stations contend simultaneously, the station possessing the highest MAC address prevails while others assume standby monitor status. All stations are designed to have the capability to become active monitors if required.

The active monitor is responsible for critical ring management tasks: it serves as the master clock to synchronize signals, inserts a 24-bit delay for buffering purposes, guarantees that only one token exists on the ring during idle periods, detects any breaks in the ring topology, and removes any frames circulating erroneously.

Token insertion process

Before a Token Ring station can join the network, it must successfully complete a five-step insertion procedure. Any failure during these steps halts the insertion and may result in driver errors.

Phase 0 (Lobe Check) – The station conducts a media integrity test by transmitting 2000 test frames through the MSAU loopback to confirm error-free reception.
Phase 1 (Physical Insertion) – The station requests the MSAU to activate its relay mechanism.
Phase 2 (Address Verification) – The station sends frames addressed to its own MAC address. If the returned frames indicate no address conflicts (both AR and FC bits equal zero), it advances to periodic ring polling for MAC address management.
Phase 3 (Participation in ring poll) – The station determines its Nearest Active Upstream Neighbour (NAUN) and announces itself downstream, aiding in ring topology mapping. It anticipates receiving AMP or SMP frames with AR and FC bits set to zero; upon receipt, it toggles these bits to one and queues an SMP frame if resources allow. Failure to receive such frames within an 18-second window leads to removal from the ring.
Phase 4 (Request Initialization) – The station obtains configuration parameters such as timers and ring identifiers from a parameter server, typically a Token Ring bridge, which are vital for proper operation.

Optional priority scheme

Certain use cases require assigning elevated priority levels to specific stations. Token Ring supports an optional priority mechanism, similar to that found in the CAN Bus—widely used in automotive networks—whereas Ethernet does not offer this capability.

The Token Ring priority MAC defines eight priority levels ranging from 0 to 7. When a station receives a token or data frame with priority less than or equal to its own requested priority, it adjusts the priority bits accordingly. The token circulates around the ring until it returns to the originating station, which then transmits data and resets the token priority back to its initial value.

The following table outlines the eight priority levels and their associated traffic categories for devices compatible with 802.1Q and 802.1p:

Priority bits	Traffic type
x'000'	Standard data traffic
x'001'	Unused
x'010'	Unused
x'011'	Unused
x'100'	Forwarded standard data traffic
x'101'	Time-critical data
x'110'	Real-time sensitive data (e.g., VoIP)
x'111'	Station management

Interconnection with Ethernet

Bridging devices facilitating communication between Token Ring and Ethernet networks included models such as the AT&T StarWAN 10:4 Bridge, IBM 8209 LAN Bridge, and Microcom LAN Bridge. Multiprotocol routers like the IBM 2210-24M integrated both Ethernet and Token Ring interfaces, enabling selective filtering of traffic types, protocols, and interfaces to support seamless network interoperability.

Operating system support

In 2012, David S. Miller applied a patch that removed Token Ring networking support from the Linux kernel, reflecting a decline in Token Ring usage within contemporary operating systems.

External links