The two-port repeater solved the fundamental problem of signal regeneration, but early Ethernet networks faced another challenge: physical topology. The original Ethernet standards (10BASE5 and 10BASE2) used a bus topology—a single coaxial cable to which all stations attached. While conceptually simple, bus topologies presented significant practical problems:

• A single cable fault could bring down the entire network
• Adding or removing stations required physical manipulation of the main cable
• Fault isolation was extremely difficult—a problem could be anywhere along the cable
• The shared medium was vulnerable to unauthorized taps

The hub emerged as the solution—a multi-port device that created a physical star topology while maintaining the logical bus behavior of Ethernet. This seemingly simple evolution represented a paradigm shift in how networks were built and managed.

A hub is fundamentally a multi-port repeater. It receives a signal on any one port and repeats that signal out all other ports simultaneously. This simple operational model conceals a sophisticated internal design that evolved significantly over the hub's commercial lifespan.

Core Definition:

A network hub is a Physical Layer (OSI Layer 1) device that:

• Provides multiple ports for connecting network devices
• Regenerates incoming signals using standard 3R techniques
• Broadcasts the regenerated signal to all ports except the originating port
• Maintains no memory of connected devices or traffic patterns
• Creates a single, extended collision domain across all ports

Internal Architecture of an Active Hub:

The internal design of a typical active Ethernet hub reveals how it achieves multi-port signal distribution:

┌─────────────────────────────────────────────────┐
│                Active Hub Internal              │
├─────────────────────────────────────────────────┤
│                                                 │
│   Port 1 ─[Transceiver]─┐                       │
│                         │                       │
│   Port 2 ─[Transceiver]─┼──[Shared Bus/Backplane]─┐
│                         │                       │ │
│   Port 3 ─[Transceiver]─┤                       │ │
│                         │    ┌──────────────┐   │ │
│   Port 4 ─[Transceiver]─┼────│ Collision    │───┘ │
│            ...          │    │ Detection    │     │
│                         │    │ & Jam Logic  │     │
│   Port N ─[Transceiver]─┘    └──────────────┘     │
│                                                 │
│   [Power Supply]    [LED Driver]   [Management] │
│                                    (optional)   │
└─────────────────────────────────────────────────┘

Key Components:

1. Port Transceivers: Each port has dedicated transmit/receive circuitry that interfaces with the connected cable, performing signal conditioning and level shifting.

2. Shared Bus/Backplane: All port transceivers connect to a common electrical bus. Any signal received on one port propagates across this bus to all other ports.

3. Collision Detection Logic: Circuitry monitors the shared bus for simultaneous transmissions. When detected, a jam signal is generated and propagated to all ports.

4. Signal Regeneration: The backplane includes the 3R circuits—re-amplification, reshaping, and retiming—applied to all signals traversing the hub.

5. Auto-Partition Logic: Many hubs include circuitry to automatically disable (partition) a port that exhibits excessive collisions or other faults, protecting the rest of the network.

The hub's most significant contribution to networking wasn't a technical innovation in signal processing—it was an architectural transformation. By converting Ethernet from a physical bus to a physical star topology, hubs revolutionized network design, installation, and maintenance.

Bus vs. Star: Physical Topology Comparison:

Centralized Wiring:

The star topology centralized all network connections at a single point—the wiring closet or telecommunications room. This physical arrangement provided numerous operational benefits:

1. Structured Cabling Systems: Star topology aligned perfectly with structured cabling standards (TIA/EIA-568), allowing permanent building infrastructure with flexible patch panel connections.

2. Fault Indication: Hub port LEDs immediately identified problematic connections. A link light that refuses to illuminate points directly to a cable or device issue.

3. Easy Moves/Adds/Changes (MACs): Moving a user required only patching a cable at the central location rather than reworking the network backbone.

4. Media Flexibility: Hubs with multiple media modules allowed mixing UTP for desktops, fiber for backbone connections, and other media types within a single logical network.

5. Physical Security: Network equipment concentrated in lockable telecommunications rooms rather than spread throughout the facility.

Logical Bus Behavior:

Despite the physical star arrangement, hubs maintained Ethernet's logical bus behavior. All devices still shared a single collision domain, and all traffic still reached all ports. This was intentional—10BASE-T (the standard that popularized hubs) was designed as a compatible evolution of 10BASE5 and 10BASE2, preserving CSMA/CD semantics while improving physical plant management.

Understanding exact hub operation is essential for comprehending its limitations and the motivations for switch development. Let's trace what happens when a frame traverses a hub.

Step-by-Step Frame Journey:

Step 1: Frame Transmission

Device A begins transmitting a frame to Device B.
The electrical signals travel through the UTP cable
to Port 3 of the hub.

Step 2: Signal Reception

The Port 3 transceiver receives the signal.
Receive circuitry performs clock recovery to synchronize
with the incoming bit stream.

Step 3: Signal Regeneration

The hub's 3R circuitry:
- Re-amplifies the signal to full voltage levels
- Re-shapes edges to eliminate distortion
- Re-times the bit stream using the recovered clock

Step 4: Bus Propagation

The regenerated signal is placed on the hub's internal bus.
All port transceivers (except Port 3) receive this signal.

Step 5: Simultaneous Transmission

Ports 1, 2, 4 through N simultaneously transmit
the regenerated signal onto their attached cables.
Every connected device receives the frame.

Step 6: Collision Handling

If Device C starts transmitting during this process:
- Both signals arrive at the hub simultaneously
- Collision detection logic recognizes the overlap
- A jam signal is generated and broadcast to all ports
- All transmitting devices back off and retry

Collision Domain Analysis:

A hub extends the collision domain to encompass all connected devices. This has quantifiable performance implications:

As the number of devices increases, collision probability rises exponentially. The theoretical maximum throughput for Ethernet with many stations under high load approaches 37% of nominal bandwidth—but practical throughput is often much lower due to variable traffic patterns and non-optimal backoff behavior.

Auto-Partitioning:

Enterprises-grade hubs implemented auto-partitioning as a fault isolation mechanism. When a port exhibits abnormal behavior (excessive collisions, jabber, etc.), the hub automatically disables that port to protect the rest of the network. The port's LED changes state to indicate the partition, and the hub periodically tests whether the condition has cleared.

Typical auto-partition triggers:

• Consecutive collision limit: More than 32-64 consecutive collisions
• Jabber detection: Continuous transmission exceeding maximum frame time
• Link pulse loss: Loss of link integrity pulses for extended period

This mechanism demonstrates that while hubs lacked Layer 2 intelligence, they incorporated sophisticated Layer 1 protection logic.

Hubs were standardized as part of the IEEE 802.3 Ethernet specifications. Understanding these standards provides insight into hub capabilities and limitations.

IEEE 802.3 Repeater Standards:

Repeater Classes (Fast Ethernet):

Class I Repeaters:

• Permit only one repeater per collision domain
• Allow translation between different physical signaling systems (e.g., 100BASE-TX to 100BASE-FX)
• Higher internal delay budget (up to 140 bit times)

Class II Repeaters:

• Permit up to two repeaters per collision domain
• Must use the same physical signaling on all ports
• Lower internal delay (up to 92 bit times)
• Faster signal propagation enables more flexibility in network design

Why No Gigabit Hubs?

The 802.3ab specification for 1000BASE-T (Gigabit Ethernet over copper) did not include a repeater specification. By the time Gigabit Ethernet was standardized in 1999, switches had become so inexpensive that there was no market for shared-medium Gigabit networks. The tight timing constraints of Gigabit Ethernet also made hub implementation impractical:

• At 10 Mbps: 1 bit time = 100 ns, minimum frame = 51.2 μs
• At 100 Mbps: 1 bit time = 10 ns, minimum frame = 5.12 μs
• At 1000 Mbps: 1 bit time = 1 ns, minimum frame = 512 ns

The nanosecond-level timing at Gigabit speeds left no practical margin for hub propagation delays while maintaining collision detection functionality.

Small networks often require more ports than a single hub provides. Connecting multiple hubs together—cascading—extends network capacity but introduces critical constraints that network designers must respect.

The 5-4-3 Rule (10 Mbps Ethernet):

For 10BASE-T networks using hubs, the 5-4-3 rule establishes maximum topology limits:

• 5: Maximum of 5 network segments (cable runs) in series between any two stations
• 4: Maximum of 4 repeaters (hubs) in the path between any two stations
• 3: Maximum of 3 populated segments (segments with stations attached)

The remaining 2 segments are IRL (Inter-Repeater Link) segments—cable runs connecting hubs to each other without attached stations.

Why These Limits Exist:

Ethernet's CSMA/CD mechanism requires that a transmitting station can detect any collision before it finishes sending the minimum-size frame (64 bytes = 512 bits). At 10 Mbps:

• Transmission time for 64 bytes: 51.2 μs
• This is the slot time—the maximum round-trip propagation delay

Signal must travel from the sender to the farthest station and back (2 × maximum distance) within 51.2 μs. Every hub adds delay:

With four hubs and five segments, the round-trip delay approaches the slot time limit.

Fast Ethernet Cascading Rules:

At 100 Mbps, timing constraints become much stricter. The slot time shrinks to 5.12 μs, allowing far less propagation distance:

Class I Repeater Networks:

• Maximum one repeater between any two stations
• Maximum collision domain diameter: ~200 meters with DTE-to-hub cables of 100m each

Class II Repeater Networks:

• Maximum two repeaters between any two stations
• Maximum 5 meters between repeaters (inter-repeater link)
• Maximum collision domain diameter: ~205 meters

These tight constraints effectively limited Fast Ethernet hub deployments to small workgroups. Larger networks required switches to create multiple collision domains.

Proper Hub Cascade Topology:

                    ┌───────────────┐
                    │   Core Hub    │
                    │   (uplinks)   │
                    └───┬───────┬───┘
                        │       │
            ┌───────────┘       └───────────┐
            │                               │
    ┌───────┴───────┐               ┌───────┴───────┐
    │  Workgroup    │               │  Workgroup    │
    │    Hub 1      │               │    Hub 2      │
    └───┬───┬───┬───┘               └───┬───┬───┬───┘
        │   │   │                       │   │   │
       PC  PC  PC                      PC  PC  PC

Note the hierarchical structure—a maximum of two hubs between any two PCs. Using switches at the core eliminated hub cascading limits entirely.

The hub's broadcast nature creates fundamental performance limitations that become more severe as network size and traffic load increase. Understanding these implications explains why hubs were ultimately replaced by switches.

Bandwidth Sharing:

A hub's bandwidth is shared among all connected devices. Unlike a switch, which can provide full bandwidth to each port simultaneously, a hub offers only aggregate bandwidth equal to a single link:

Collision Overhead:

As more devices share the collision domain, collision probability increases non-linearly. The mathematical analysis shows:

Ethernet throughput under load (S) is approximately:

S = G × e^(-2G)

Where G = offered load (attempts per slot time)
Maximum throughput occurs at G = 0.5, giving S ≈ 0.18 (18%)

This theoretical maximum assumes ideal backoff behavior. Practical measurements show:

• Light load (<30% utilization): Minimal collisions, good performance
• Moderate load (30-60%): Increasing collisions, noticeable delays
• Heavy load (>60%): Collision storms, severe throughput degradation
• Saturation (>80%): Network approaching unusability

Half-Duplex Limitation:

Hubs inherently operate in half-duplex mode—only one device can transmit at a time across the entire hub. This is fundamental to the collision domain concept. Full-duplex Ethernet, where devices can transmit and receive simultaneously, requires switches (or direct point-to-point connections).

Impact on Modern Applications:

Even legacy hubs would struggle with modern network traffic patterns:

Latency Variability:

Unlike switches with predictable store-and-forward delays, hub latency varies wildly based on collision patterns:

• Best case: ~10 μs (signal regeneration only)
• Typical case: ~100 μs (one collision, successful retry)
• Worst case: ~100+ ms (multiple collision cascade)

The hub's broadcast nature creates significant security vulnerabilities that modern network designs must avoid. Understanding these vulnerabilities emphasizes why hubs are inappropriate for any security-conscious environment.

Passive Eavesdropping:

Because every frame transmitted on a hub reaches every port, any device can capture all network traffic simply by placing its network interface in promiscuous mode. No special hardware or sophisticated attack is required—standard packet capture tools (Wireshark, tcpdump) work directly.

This means on a hub-based network:

• All passwords transmitted in clear text (Telnet, early HTTP, FTP) are visible to all users
• Email content traversing the network can be read by any connected device
• Proprietary application traffic can be analyzed and reverse-engineered
• Session cookies can be captured for session hijacking

No Access Control:

Hubs provide no mechanism for:

• Port-based authentication (802.1X is impossible)
• MAC address filtering (hubs don't see MAC addresses)
• VLAN segmentation (hubs can't tag or recognize tags)
• Traffic isolation of any kind

Regulatory Implications:

Modern security and privacy regulations make hub deployments problematic or impossible:

The Security Case for Switches:

Switches fundamentally change the security model:

• Traffic is forwarded only to the intended destination port
• MAC address tables limit traffic visibility
• Port security can restrict which devices may connect
• 802.1X can require authentication before network access
• VLANs provide logical segmentation
• Traffic monitoring requires specific port mirroring configuration

The transition from hubs to switches represents one of the most complete technology replacements in networking history. Understanding why this happened illuminates the fundamental limitations of shared-medium networking.

Economic Crossover:

In the early 1990s, switches cost 10-20x more than hubs on a per-port basis. A 24-port 10BASE-T hub might cost $1,000; an equivalent switch cost $10,000-$20,000. This price premium initially restricted switches to backbone connections.

By the mid-2000s, the economics inverted:

Once switch prices fell to near-hub levels, the hub had no remaining advantage. Manufacturers stopped production; retailers cleared inventory; new installations universally specified switches.

Where Hubs Might Still Exist:

Despite their obsolescence, hubs occasionally appear in specific scenarios:

1. Network Analysis: Security professionals sometimes use hubs to create monitoring points—unlike switches that require port mirroring configuration, hubs naturally provide all traffic to capture devices.

2. Legacy Equipment: Some old industrial or medical equipment requires hub behavior for protocol reasons or has embedded hubs in aged networking components.

3. Very Old Installations: Networks in abandoned buildings, historical systems preserved for documentation, or extremely neglected infrastructure.

4. Education: Networking courses sometimes use hubs to demonstrate collision behavior and contrast with switch operation.

Current Availability:

New hubs are essentially unavailable from mainstream vendors. Remaining inventory is typically:

• Used/refurbished units
• Specialty industrial versions
• Educational lab equipment
• Network testing tools

The hub era ended sometime around 2005-2010, completing a technology transition that began in the mid-1990s.

We have comprehensively explored the network hub—the multi-port repeater that transformed physical network topology while maintaining shared-medium behavior. Let's consolidate the key concepts:

Looking Ahead:

The hub represented an important evolutionary step—solving the physical topology problems of bus networks while maintaining Ethernet semantics. However, its shared-medium nature created inherent limitations in performance, scalability, and security.

In the next page, we'll explore the bridge—the first network device to operate at Layer 2 (Data Link Layer) and provide collision domain segmentation. This conceptual leap from signal repetition to intelligent frame forwarding laid the groundwork for modern switch technology.