Loading learning content...
Network interconnection devices are the architectural building blocks that define network topology, performance characteristics, and segmentation boundaries. Each device type—from simple repeaters to sophisticated routers—interacts differently with collision and broadcast domains based on its layer of operation and functional design.
Understanding these interactions is not merely academic. It directly informs design decisions: Should I use a hub or a switch? How many broadcast domains will this design create? Why is performance degrading after adding that network segment? The answers lie in understanding how devices shape domain boundaries.
A fundamental principle governs device behavior: the higher the OSI layer a device operates at, the more segmentation it provides. Layer 1 devices provide no segmentation. Layer 2 devices segment collision domains only. Layer 3+ devices segment both collision and broadcast domains. This principle will guide our entire analysis.
A repeater is the simplest network interconnection device. Operating entirely at the Physical Layer (Layer 1), a repeater receives electrical signals on one port, amplifies and regenerates them, and retransmits them on another port. The repeater has no awareness of frames, addresses, or any higher-layer concepts—it deals purely with signal waveforms.
Physical Layer Operation
When a signal arrives at a repeater:
This process happens at wire speed with minimal latency (typically < 1 microsecond).
Why Repeaters Are Needed
Signals attenuate (weaken) and distort as they travel along cables:
Without repeaters, the maximum cable length would be severely limited. Repeaters extend reach by regenerating the signal before it degrades beyond recognition.
| Characteristic | Description | Impact |
|---|---|---|
| OSI Layer | Layer 1 (Physical) | No awareness of frames or addresses |
| Collision Domain | Extends the collision domain | All connected segments become one collision domain |
| Broadcast Domain | Extends the broadcast domain | Broadcasts propagate through the repeater |
| Latency | Minimal (< 1 µs typical) | Negligible delay added |
| Collision Handling | Propagates collisions | Collision detected on one side affects both sides |
| Segmentation | None | Does not reduce contention or isolate traffic |
Domain Impact Analysis
When you connect two network segments with a repeater:
[Segment A] ----[Repeater]---- [Segment B]
The 5-4-3 Rule
Traditional Ethernet specifications limited repeater use to prevent excessive latency (which would break CSMA/CD timing):
This rule ensured the round-trip delay remained within the slot time budget.
Modern networks rarely use standalone repeaters. Their function has been absorbed by hubs (multiport repeaters) and made largely irrelevant by switches. You're most likely to encounter repeater concepts in legacy networks, certification exams, or specialized applications like long-distance fiber links with optical repeaters.
A hub is essentially a multiport repeater. It operates at the Physical Layer and connects multiple devices to a shared collision domain. Any signal received on one port is regenerated and transmitted out all other ports—the hub makes no decisions about where to send traffic.
Hub Architecture
Internally, a hub consists of:
When any port receives a signal:
Domain Impact Analysis
A hub creates a single collision domain containing all connected devices and a single broadcast domain encompassing the same devices.
Example: An 8-port hub with 8 connected workstations:
Cascaded Hubs
Connecting hubs together extends the collision domain:
[Hub A]----[Hub B]----[Hub C]
| | |
PCs PCs PCs
In modern networks, there are very few valid reasons to use a hub: (1) Network diagnostics where you need to capture all traffic without installing port mirroring, (2) Extremely cost-sensitive applications with low traffic and few nodes, (3) Legacy equipment compatibility. In virtually all other cases, switches are superior and often cheaper.
A bridge is a Layer 2 device that segments collision domains while maintaining a single broadcast domain. Bridges revolutionized network design by providing intelligent traffic forwarding based on MAC addresses—a fundamental advancement over the dumb signal repetition of hubs and repeaters.
Bridge Operation
Bridges employ a store-and-forward methodology:
Learning and Aging
The MAC address table (also called CAM table, SAT, or forwarding database) is built dynamically:
| Characteristic | Description | Impact |
|---|---|---|
| OSI Layer | Layer 2 (Data Link) | Understands MAC addresses; makes forwarding decisions |
| Collision Domain | Separates collision domains | Each port is an independent collision domain |
| Broadcast Domain | Extends the broadcast domain | Broadcasts are flooded to all ports |
| Processing | Store-and-forward | Adds latency (time to receive full frame) |
| Intelligence | MAC learning + filtering | Reduces unnecessary traffic on each segment |
| Typical Port Count | 2-4 ports | Limited scalability; switches are preferred |
Domain Impact Analysis
Consider a 2-port bridge connecting two hub-based network segments:
[Hub A (5 PCs)]----[Bridge]----[Hub B (5 PCs)]
Before the bridge (if hubs were directly connected):
After adding the bridge:
Traffic Filtering Example
If PC1 on Hub A sends to PC2 (also on Hub A):
This filtering is a major performance improvement over hubs—traffic that doesn't need to cross the bridge is contained to its local segment.
Early bridges were software-based, processing frames in a CPU. This limited their throughput and port count. The term 'switch' emerged when hardware-based (ASIC) bridging became possible, enabling wire-speed forwarding and higher port densities. Technically, a switch is a multiport bridge with hardware acceleration. The terms are often used interchangeably today.
A switch is a multiport bridge implemented in hardware for wire-speed operation. Switches fundamentally transformed Ethernet by providing microsegmentation—each port becomes its own collision domain. Combined with full-duplex support, switches effectively eliminated the collision-related throughput ceiling that limited shared Ethernet.
Switch Architecture
Modern switches use specialized hardware for high-performance forwarding:
Forwarding Methods
Switches may use different forwarding strategies:
| Method | Operation | Latency | Error Handling |
|---|---|---|---|
| Store-and-Forward | Receive entire frame, validate FCS, then forward | Highest (full frame time) | Drops corrupted frames |
| Cut-Through | Start forwarding after reading destination MAC | Lowest (~6 bytes delay) | May forward corrupted frames |
| Fragment-Free | Forward after receiving first 64 bytes | Medium (~64 bytes delay) | Catches runt frames (collision fragments) |
| Adaptive | Cut-through normally; falls back to store-and-forward if error rate rises | Variable | Best of both worlds |
Domain Impact: Microsegmentation
A switch provides one collision domain per port. This microsegmentation has profound implications:
Example: 24-port switch with 24 connected workstations:
Full-Duplex Operation
When a switch port connects to a single device (not a hub), full-duplex operation becomes possible:
Switch vs. Hub Performance Comparison
Consider 10 hosts generating 50 Mbps of traffic each (100 Mbps network):
| Metric | 10-Port Hub | 10-Port Switch (Full-Duplex) |
|---|---|---|
| Collision Domains | 1 | 10 |
| Total Bandwidth | 100 Mbps shared | 1,000 Mbps aggregate (100 Mbps × 10) |
| Per-Host Bandwidth | ~10 Mbps average | 100 Mbps dedicated |
| Collisions | Frequent (high contention) | None (full-duplex) |
| Latency | Variable (collision backoff) | Consistent (< 10 µs typical) |
| Security | Poor (all traffic visible) | Better (unicast only to destination port) |
Without VLANs, all switch ports share one broadcast domain. VLANs add Layer 2 segmentation: each VLAN is a separate broadcast domain. A 24-port switch could have ports 1-8 in VLAN 10, ports 9-16 in VLAN 20, and ports 17-24 in VLAN 30—creating three broadcast domains on a single physical switch. Inter-VLAN communication requires a router or Layer 3 switch.
A router operates at the Network Layer (Layer 3), making forwarding decisions based on IP addresses rather than MAC addresses. Each router interface connects to a separate network and—critically—a separate broadcast domain. Routers provide the strongest domain separation and are essential for building scalable, hierarchical network architectures.
Router Operation
When a packet arrives at a router:
Why Routers Stop Broadcasts
Routers do not forward broadcast frames for fundamental design reasons:
Domain Impact Analysis
Consider a router with 3 interfaces:
[Network A: 10.0.1.0/24]----[Router]----[Network B: 10.0.2.0/24]
|
[Network C: 10.0.3.0/24]
Domain Boundaries:
Directed Broadcast Handling
A directed broadcast (e.g., 10.0.2.255 targeting Network B) could theoretically be converted to a Layer 2 broadcast on the destination network. However:
DHCP and Broadcast Boundaries
When broadcasts can't cross routers, protocols that depend on broadcasts need special handling:
Layer 3 switches combine switching and routing in a single device. They switch traffic within a VLAN at wire speed (Layer 2) and route between VLANs using hardware-accelerated routing (Layer 3). This makes them the preferred choice for enterprise networks: VLAN switching performance with router-like broadcast domain separation.
Let's consolidate all device behaviors into a comprehensive comparison that can serve as a reference for network design and troubleshooting.
| Device | OSI Layer | Collision Domain Effect | Broadcast Domain Effect | Forwarding Basis |
|---|---|---|---|---|
| Repeater | Layer 1 | Extends ↔ | Extends ↔ | Electrical signals (all) |
| Hub | Layer 1 | Extends ↔ | Extends ↔ | Electrical signals (all) |
| Bridge | Layer 2 | Terminates ✓ | Extends ↔ | MAC addresses |
| Switch (no VLANs) | Layer 2 | Terminates ✓ | Extends ↔ | MAC addresses |
| Switch (with VLANs) | Layer 2 | Terminates ✓ | Terminates ✓ (per VLAN) | MAC + VLAN tag |
| Router | Layer 3 | Terminates ✓ | Terminates ✓ | IP addresses |
| Layer 3 Switch | Layer 2/3 | Terminates ✓ | Terminates ✓ (per VLAN) | MAC or IP (context) |
| Firewall | Layer 3+ | Terminates ✓ | Terminates ✓ | IP + policy rules |
Quick Reference Rules
Counting Collision Domains:
Collision Domains = (Switch Ports) + (Router Interfaces) + (Bridge Ports)
Note: A hub-connected segment counts as 1 collision domain, regardless of hosts.
Counting Broadcast Domains:
Broadcast Domains = (Router Interfaces) + (Number of VLANs)
Note: Each VLAN or router-separated network is one broadcast domain.
Example Network Analysis:
Collision Domains: (4 × 24 switch ports) + (2 × 3 router interfaces) + 2 (one per hub) = 96 + 6 + 2 = 104
Broadcast Domains: 6 router interfaces = 6 (or fewer if some interfaces connect to same segment)
Modern network design favors Layer 2 switches with VLANs for access/distribution layers (provides microsegmentation and broadcast control) and Layer 3 switches/routers for core/distribution layers (provides routing and definitive broadcast domain separation). Hubs are essentially obsolete, and standalone bridges are rare outside specialized applications.
We have systematically analyzed how each major network interconnection device affects collision and broadcast domain boundaries. This knowledge forms the foundation for network design and troubleshooting.
With a thorough understanding of how devices impact domains, you're prepared to learn about domain sizing strategies. The next page examines how to determine optimal broadcast domain sizes for different network scenarios, balancing performance, scalability, and administrative concerns.