Network Layer (Routing and Routing Tables)
Context: FIT1047_MOC · how packets cross the network-of-networks — table lookups hop by hop · subnet membership decided in Network Addresses (URL, Port, IP, MAC)
Quick Revision
- 🎯 Objective: per packet: read destination IP → routing table → next router (or direct delivery) → else default gateway → else unroutable.
- 📦 Core Components: subnets ➔ routers (layer-3, one IP per interface) ➔ routing tables ➔ static vs dynamic (distance vector vs link state).
- ⚡ Key Constraint: distance vector (share routing tables, pick fewest hops — RIP) vs link state (share connectivity/link quality, pick fastest path — OSPF).
📝 Core
- Subnet recap ➔ all devices directly connected, sharing an address prefix ( etc.); in-subnet traffic goes direct, anything else goes to a router.
- Routers ➔ THE Internet infrastructure: layer-3 devices, one IP address per interface (≈ per connected subnet); forward by destination IP.
- Table lookup per packet ➔ destination in table? → forward to listed next router or deliver directly · not in table? → default gateway · no gateway? → packet unroutable.
- Static routing ➔ tables hand-written by the network manager; fine for small, stable networks.
- Dynamic routing ➔ routers exchange information and converge on tables themselves.
- Distance vector ➔ exchange distance-to-destination info, choose SHORTEST (hop count): RIP (max 15 hops — loop guard; whole-table updates; slow convergence ⟹ small networks), EIGRP, BGP.
- Link state ➔ exchange full connectivity + link-quality info, choose FASTEST route by metric (usually link speed): OSPF.
📊 Exam Execution Trace — lecture topology (routers A…K)
A packet for H arriving at router with table {H → G}: forwarded to G; G’s table {H → H}: delivered. A destination absent from the table follows default → C; if C also lacks both entry and gateway, the packet dies. Method: at every router, ONE lookup, ONE hop decision — never a full path plan.
⚖️ Core Decision Matrix
| distance vector | link state | |
|---|---|---|
| exchanges | its routing table (distances) | connectivity + link quality of known network |
| optimises | fewest hops | fastest links |
| protocols | RIP (≤15 hops), EIGRP, BGP | OSPF |
| scale | small networks (RIP) | larger, faster convergence |
⚠️ Common Mistakes
- 💡 Routers plan one hop, not the path ➔ each router only knows “next router for this destination”; end-to-end routes emerge, they aren’t computed anywhere.
- 💡 RIP’s 15-hop cap is a feature ➔ it bounds count-to-infinity loops, and simultaneously confines RIP to small networks.
- 💡 Shortest ≠ fastest ➔ two hops over fibre beat one hop over ADSL; that’s the whole link-state argument.
🧠 Active Recall
A router receives a packet for an IP absent from its table. Walk the decision chain.
Answer
- Short answer: Table miss → forward to the configured default gateway; no default gateway → the packet cannot be routed (dropped).
- Why: Default as aggregate ➔ the gateway entry stands for “everything I don’t know” — the reason edge routers survive with tiny tables.
Contrast what RIP and OSPF routers actually SEND each other, and what each optimises.
Answer
- Short answer: RIP sends its whole routing table (distances in hops) → shortest hop-count wins; OSPF sends link-state (who’s connected, link speeds) → fastest metric path wins.
- Why: Information richness ➔ knowing the map (link state) lets each router compute better routes than knowing neighbours’ summaries (distance vector).