Telecommunications Fundamentals and Network Architecture
This domain tests the structural vocabulary every other domain assumes: layering, switching modes, the access–metro–core hierarchy, and how redundancy is quantified. Questions typically ask you to place a device at the correct layer, contrast circuit and packet behaviour, and compute end-to-end availability.
Layering and planes
- OSI seven layers versus the four-layer TCP/IP model. Device placement is examinable: hub/repeater L1, switch/bridge L2, router L3, firewall L3–L7, load balancer L4–L7.
- Three planes: control (signalling, routing), user/bearer (payload), management. SDN later relocates control functions off the forwarding element.
- Circuit switching reserves capacity for the whole session: constant rate, low jitter, wasteful for bursty data. Packet switching statistically multiplexes: efficient, but needs buffering, queueing discipline and QoS marking.
Topologies and hierarchy
- Full mesh maximises path diversity but scales as n(n−1)/2 links; partial mesh is the practical compromise. Star/hub-and-spoke centralises cost and creates a single point of failure; rings are typical in metro transport where protection switching is required.
- Traffic hierarchy: access (last mile, per-subscriber) → metro/aggregation (grooming, concentration) → core/backbone (long haul, high capacity). Each level aggregates and grooms, so capacity ratios between levels are a design rule, not an accident.
- Interconnection: peering is usually settlement-free at an internet exchange; transit is paid and delivers the full routing table. Content delivery caches push traffic closer to the edge and reduce transit cost.
Worked example
Two parallel core links each 99.9% available: 1 − (1 − 0.999)² = 1 − 0.000001 = 99.9999%. If a single access link at 99.9% sits in series with that core, end-to-end availability = 0.999 × 0.999999 ≈ 99.90%. The lesson: redundancy added where availability is already high buys almost nothing; the series element with the lowest availability dominates the result and must be fixed first.
Traps that cost marks
- Calling a Layer 2 switch a router, or claiming a firewall operates only at Layer 3.
- Believing parallel redundancy multiplies availabilities; it improves them (product of failure probabilities).
- Treating bandwidth as latency: a 1 Gbit/s link can still have 60 ms delay if propagation and queueing are high.
- Assuming full mesh is always the best design - cost and port count usually rule it out.
- Forgetting that availability is a probability over time, not a measure of error rate.