Topology Comparison

Part of Networking

Comparing network topologies on cost, performance, reliability, and scalability.

Why This Matters

Choosing the right network topology is one of the most consequential early decisions in network design. The topology determines cable quantities, hardware requirements, failure behavior, and maintainability — all of which have long-term implications. Changing topology after a network is built and cabled is extremely expensive, so getting it right from the start matters.

Understanding topology tradeoffs goes beyond knowing which topology is “best.” Different contexts genuinely benefit from different topologies. A factory floor with equipment that cannot tolerate any network disruption might need a ring topology with automatic failover. A small office needs a simple star at minimal cost. A large campus needs a hierarchical network combining multiple topologies.

This article provides a systematic comparison of the main network topologies across the dimensions that matter most in practice.

Bus Topology

In bus topology, all devices share a single cable backbone. Any transmission propagates to all devices.

Cost: Low initial cost — a single cable serves all devices, and each connection needs only a T-connector and short drop cable. Total cable quantity is minimal.

Performance: Decreases as device count increases. All devices share one collision domain. CSMA/CD manages access but performance degrades with load. Effective throughput per device roughly halves each time device count doubles.

Reliability: Poor. Any break in the backbone cable, any missing terminator, or any malfunctioning device that continuously transmits can disrupt the entire network. Diagnosing the fault location requires physical inspection of the entire cable.

Scalability: Poor. Beyond 20-30 devices, collision rates become problematic. Maximum cable length is 185 meters for 10Base2.

Maintenance: Difficult. Adding a device requires interrupting the cable. Tracing a fault requires inspecting the entire backbone.

Best for: Historical curiosity. Not recommended for any new installation. Understand bus topology to deal with legacy equipment, not to build new networks.

Star Topology

All devices connect individually to a central hub or switch.

Cost: Moderate. More cable than bus (separate run to each device), plus the cost of the central switch. Cable cost scales linearly with device count.

Performance: Excellent with switches. Each port gets its own collision domain; full-duplex switches eliminate collisions entirely. Performance is independent of device count (within switch capacity).

Reliability: Good. Faults are isolated to individual links. The central switch is a single point of failure, but switches rarely fail; UPS protection and redundant switches address this.

Scalability: Excellent for moderate sizes. Hierarchy (switches connected to switches) extends star topology to thousands of devices.

Maintenance: Easy. Each link is independent. Adding a device requires only plugging into an available switch port. Diagnosing a fault requires checking only the suspect device’s cable and switch port.

Best for: Nearly all new installations. Default choice unless specific requirements argue for another topology.

Ring Topology

Devices form a closed loop; each device connects to two neighbors.

Cost: Moderate. Cable quantity similar to star. Requires ring-capable network equipment.

Performance: Moderate. Token Ring manages access with no collisions. Performance is predictable but lower peak throughput than switched Ethernet.

Reliability: Good for self-healing rings (bidirectional ring with automatic failover). A single link failure reroutes traffic on the alternate path. Poor for single-path rings (one failure disrupts the entire ring).

Scalability: Limited. Token Ring networks are limited in size by token circulation latency.

Maintenance: Moderate. Adding a device requires inserting it into the ring, which momentarily disrupts the existing ring.

Best for: Applications requiring deterministic access and self-healing (industrial fiber rings, metropolitan area networks).

Mesh Topology

Nodes are interconnected with multiple paths between them.

Cost: High. Number of links grows quadratically with node count for full mesh. Even partial mesh requires careful planning to justify cost.

Performance: Excellent. Multiple paths allow load distribution and parallel forwarding. No single bottleneck link.

Reliability: Excellent. Multiple paths mean multiple failures required to disconnect any node. Routing protocols automatically use alternate paths.

Scalability: Limited for full mesh (impractical beyond 5-10 nodes). Partial mesh scales to large networks when combined with hierarchical design.

Maintenance: Complex. Many connections to track and manage.

Best for: Network backbone (connecting core routers), data center fabrics, and any scenario where high availability justifies the cost.

Hierarchical (Tree) Topology

Combines multiple levels of star topology — access layer connects end devices, distribution layer connects access switches, core layer connects distribution switches.

Cost: Scales well. Access layer uses inexpensive switches; higher layers use progressively more capable (and expensive) equipment.

Performance: Excellent if properly designed. Uplinks between layers are sized to handle aggregate traffic from below. Bottlenecks occur if uplinks are undersized.

Reliability: Variable depending on redundancy built into each layer. Core layer should be redundant; access layer typically is not.

Scalability: Excellent. Hierarchy scales to thousands or millions of devices (the internet is a hierarchical network).

Maintenance: Manageable if well-documented. Changes at one layer do not necessarily affect other layers.

Best for: Any network larger than a small office. The default architecture for buildings, campuses, and enterprises.

Quantitative Comparison Table

Property	Bus	Star/Hub	Star/Switch	Ring	Mesh	Hierarchical
Cable quantity	Low	Medium	Medium	Medium	High	Medium-High
Hardware cost	Low	Low-Med	Medium	Medium	High	Medium-High
Fault isolation	None	Per link	Per link	Per link	Per path	Per segment
Collision domain	Whole bus	Whole hub	Per port	None	N/A	Per access segment
Failure tolerance	None	Low	Low	Self-heal (dual ring)	High	Depends on design
Max devices	~30	~8-16	Hundreds	~40 (TR)	~10 (full)	Thousands
New installation?	No	No	Yes	Special	Special	Yes (large)

Decision Framework

When choosing a topology, work through these questions in order:

1. How many devices? Fewer than 10: simple star. Dozens to hundreds: star with hierarchy. Thousands: full hierarchical with redundant core.

2. What is the cost constraint? Minimal budget favors simple star with unmanaged switch. Moderate budget allows managed switches and some hierarchy. Larger budget enables redundancy and mesh for critical links.

3. What is the reliability requirement? Tolerable downtime: simple star, single switch. Minimal downtime: redundant switches, dual-homed critical devices. Near-zero downtime: mesh core, multiple uplinks, redundant power.

4. Are there special requirements? Industrial real-time control may need deterministic access (ring). Very long distances need fiber, possibly ring or point-to-point links. Wireless coverage needs star with access points at the right locations.

5. What must the network evolve to? A network that will grow by 10x in five years needs a more scalable initial design than one that will stay the same size. Design for the future size, not just the present.

For most situations — office networks, small facilities, modest budgets — the answer is switched star with hierarchical extension as needed. This topology requires no special justification; it is the obvious choice. Only when specific requirements cannot be met by switched star should you consider alternatives.