Token Ring
IBM's Token Ring was the main competitor to Ethernet in the 1980s and 1990s.
How Token Passing Works
Unlike Ethernet's CSMA/CD where any device can transmit when the line is idle (and collisions must be detected and resolved), Token Ring uses a special 3-byte frame called a token that circulates continuously around a logical ring. Only the station holding the free token may transmit data. After transmitting, the station releases the token for the next station in the ring.
The ring is a logical loop — physically, cables run from station to station through a MAU (Multistation Access Unit) which provides the ring topology electrically. Each station regenerates and forwards every frame, acting as a repeater. The token passes from station to station with a maximum propagation delay determined by the ring latency.
The Token Frame
The token is a 3-byte frame consisting of:
- SD (Start Delimiter): 1 byte — 10101011 with deliberate Manchester encoding violations, signaling the start of a frame
- AC (Access Control): 1 byte — contains priority (3 bits), reservation (3 bits), token bit (0 = free, 1 = busy), and monitor bit
- ED (End Delimiter): 1 byte — 10101111 with intermediate-frame and error-detection bits
When a station wants to transmit, it seizes the free token by setting the token bit to 1 (busy), appends its data frame, and transmits the complete frame around the ring. Each station examines the destination address, copies the frame if addressed to it, and regenerates the signal for the next station. When the frame returns to the originating station, it is removed from the ring and the station releases a new free token.
Token Release Modes
- Early Token Release (ETR): The transmitting station releases the token immediately after transmitting its frame, rather than waiting for the frame to circle the entire ring. This doubles effective throughput at low utilization. Required for 16 Mbps operation.
- Traditional release: The station waits for its own frame to return (confirming successful transmission) before releasing the token. Guarantees delivery confirmation but reduces throughput.
Ring Topology and the MAU
A Token Ring network uses a MAU (Multistation Access Unit) — a wiring hub that contains the ring topology internally. Each station connects to the MAU via a dedicated cable. The MAU contains bypass relays that can electrically remove a station from the ring if it fails or is powered off, maintaining ring continuity without manual intervention.
Multiple MAUs can be daisy-chained to expand the ring. The standard limits the ring to 260 stations per ring (with a practical limit of ~72 for performance). Maximum cable length depends on the medium:
- Shielded Twisted Pair (STP): 100m between stations, 360m total ring length
- Unshielded Twisted Pair (UTP, Cat 3+): 100m between stations, 360m total
- Fiber optic: 2km between stations, used for backbone segments
Active Monitor and Ring Maintenance
One station on the ring is elected Active Monitor — responsible for ring maintenance. The Active Monitor:
- Generates the ring beacon: If the ring is broken, the Active Monitor sends beacon frames to identify the fault location
- Clocks the ring: Provides the master clock for all stations on the ring
- Purge stale frames: Removes frames that have circulated too long (preventing infinite loops)
- Manages ring polling: Sends periodic polling frames to verify ring health
All other stations operate as Standby Monitors, watching for beacon frames and the Active Monitor's polling. If the Active Monitor fails, another station is elected via a monitoring competition process. The monitor bit in the AC byte is used to track whether a frame has passed through the Active Monitor.
Priority Access Mechanism
Token Ring supports 8 priority levels (0–7), with 0 being the lowest and 7 the highest. This makes Token Ring deterministic — critical for real-time applications like industrial control.
When a station wants to transmit at a priority above 0, it sets thereservation field in the AC byte of any passing frame or token. When the current transmission completes and a free token is released, the station with the highest reservation value wins the token. Stations at lower priorities must wait.
The Token Holding Timer (THT) limits how long a station can hold the token. At 4 Mbps, the default THT is 10 ms. At 16 Mbps, it is 1.4 ms. After the THT expires, the station must release the token even if it has more data to send, ensuring fair access across all stations.
IEEE 802.5 Frame Format
Token Ring data frames follow this structure:
SD | AC | FC | DA (6 bytes) | SA (6 bytes) | INFO (up to 17,800 bytes) | FCS (4 bytes) | ED | FS
- SD (Start Delimiter): 1 byte — 10101011 with Manchester encoding violations
- AC (Access Control): 1 byte — priority, reservation, token, monitor
- FC (Frame Control): 1 byte — frame type (LLC, MAC, or beacon)
- DA (Destination Address): 6 bytes — individual or group address
- SA (Source Address): 6 bytes — always an individual address
- INFO: Up to 17,800 bytes of payload (LLC/MAC data)
- FCS (Frame Check Sequence): 4 bytes — CRC-32 for error detection
- ED (End Delimiter): 1 byte — 10101111 with I (intermediate) and E (error) bits
- FS (Frame Status): 1 byte — address-recognized and frame-copied bits
The maximum payload of 17,800 bytes means Token Ring frames are significantly larger than Ethernet's 1,500 byte MTU. This reduces per-frame overhead for large transfers but increases latency for small frames.
Beaconing and Fault Recovery
When a station detects a serious ring fault (no signal, excessive errors), it begins sending beacon frames— continuous transmissions containing the station's own address. This serves two purposes:
- Fault isolation: The beacon source address identifies where the ring break is located
- Automatic recovery: The MAU can activate its bypass relay around the beaconing station
After bypassing the faulty station, the Active Monitor initiates aring purge — sending a series of frames to clear any residual frames and re-synchronize the ring clock. Recovery typically completes in under 1 second.
IBM Cabling Specifications
IBM defined specific cable types for Token Ring, each with distinct characteristics:
| Type | Impedance | Shielding | Data Rate | Application |
|---|---|---|---|---|
| Type 1 | 150 Ω | STP (braided shield) | 4/16 Mbps | Main backbone, workstation |
| Type 2 | 150 Ω | STP + 2 telephone pairs | 4/16 Mbps | Combined voice/data |
| Type 3 | 100 Ω | UTP (no shield) | 4 Mbps only | Low-cost installations |
| Type 5 | — | Fiber optic | 4/16 Mbps | Long-distance backbone |
| Type 6 | 150 Ω | STP (lightweight) | 4/16 Mbps | Workstation patch cables |
The 150 Ω impedance of Type 1 STP cable is unique to Token Ring — it does not match the 100 Ω impedance used by Ethernet. This incompatibility meant Token Ring required dedicated cabling infrastructure, increasing deployment cost.
Token Ring vs Ethernet: Technical Comparison
| Feature | Token Ring (802.5) | Ethernet (802.3) |
|---|---|---|
| Access method | Token passing (deterministic) | CSMA/CD (contention-based) |
| Collisions | None | Common under load |
| Max data rate | 16 Mbps (HSTR: 100 Mbps) | 10 Mbps → 100 Mbps → 1 Gbps |
| Max frame size | 17,800 bytes | 1,500 bytes |
| Topology | Logical ring (via MAU) | Bus or star |
| Max stations | 260 per ring | Unlimited (switch-based) |
| Cable cost | High (150 Ω STP required) | Low (100 Ω UTP Cat 3+) |
| Under load | Consistent latency | Exponential backoff degrades |
| Fault tolerance | MAU bypass relays | Switch port isolation |
Why Token Ring Lost to Ethernet
Token Ring was technically superior in several ways — deterministic latency, no collisions, built-in priority, and automatic fault recovery. But Ethernet won the standards war for economic and architectural reasons:
- Cost: Token Ring adapters cost $400–$800 in 1990 dollars, while Ethernet adapters fell below $100. MAUs cost $500–$2,000 versus $50 for an Ethernet hub.
- Speed progression: Fast Ethernet (100 Mbps) arrived in 1995 at a fraction of HSTR's cost. Gigabit Ethernet arrived in 1999. Token Ring stalled at 16 Mbps for most deployments.
- Cabling: Token Ring required expensive 150 Ω STP cabling. Ethernet worked on the same cheap UTP cable used for telephone wiring.
- Switching: Ethernet switched from shared-bus hubs to full-duplex switches, eliminating collisions entirely and making Token Ring's collision-free advantage irrelevant.
- Ecosystem: Ethernet had massive economies of scale — more vendors, more tools, more expertise. Token Ring remained an IBM-centric ecosystem.
- IBM's retreat: When IBM de-emphasized Token Ring in the mid-1990s, the market collapsed. Cisco, the last major Token Ring vendor, discontinued products by 2004.
The End of Token Ring
By 2000, Ethernet's 100Base-TX dominated LANs worldwide. IBM exited the Token Ring business in 2000, selling technology to Cisco and HP. IEEE formally withdrew the 802.5 standard in 2004. By 2010, Token Ring was essentially extinct — though a handful of legacy systems persisted in industrial environments (steel mills, power plants) where deterministic performance and backward compatibility outweighed the cost of maintaining obsolete hardware.
Token Ring's legacy lives on in the concept of deterministic networking — the same principles that made Token Ring attractive for real-time applications are now implemented in Time-Sensitive Networking (TSN) on Ethernet, bringing deterministic guarantees to the very standard that killed Token Ring.