Network Topologies

Topology is the geometry of connection. It defines how nodes (computers) and links (cables) are arranged. This decision impacts cost, reliability, and latency.

There are two types:

  • Physical Topology: How the cables are actually laid out.
  • Logical Topology: How data flows (e.g., a physical Star can behave like a logical Bus).

1. Classic Topologies

Topology Description Pros Cons
Star All nodes connect to a central Switch. Easy to fix. If a cable breaks, only 1 PC dies. If the Switch dies, everything dies.
Mesh Every node connects to every other node. Zero Downtime. If a link breaks, traffic reroutes. Expensive. Requires N(N-1)/2 cables.
Ring Nodes form a loop. Tokens pass around. Deterministic latency (Token Ring). One break kills the loop (unless dual-ring).
Bus Single backbone cable. Cheap and simple. Collision hell. One cut splits the network.

2. Modern Architecture: Spine-Leaf (Clos Network)

In modern Data Centers (like Google/AWS), the classic “Three-Tier” (Access-Agg-Core) architecture creates bottlenecks. We now use Spine-Leaf.

  • Leaf Switches: Connect to Servers.
  • Spine Switches: Connect to Leaf Switches.
  • Rule: Every Leaf connects to EVERY Spine. No Spine connects to another Spine.

[!TIP] Why Spine-Leaf? It guarantees deterministic latency. Any server is exactly 3 hops away from any other server in the data center (Leaf → Spine → Leaf). It also allows massive East-West traffic flow, which is critical for distributed systems like Kubernetes.

3. Interactive: Reliability Simulator

See how different topologies handle failure. Click “Cut Link” to simulate a cable break.

Status: Healthy

4. Reliability Math

How much more reliable is a Mesh? Let’s use probability.

  • Let p be the probability that a single link fails (e.g., 0.1 or 10%).
  • Star: If the link to Node A fails, Node A is down. P(Failure) = p.
  • Mesh: To isolate Node A, all its links must fail simultaneously. If Node A has 4 links:
  • P(Failure) = p × p × p × p = p4
  • If p=0.1, Star failure is 10%. Mesh failure is 0.01% (1 in 10,000).

[!NOTE] This is why critical systems (nuclear plants, aircraft) use triple-modular redundancy. The math makes simultaneous failure statistically impossible.