Advanced Networking

Part 1: Core Concepts (2 Marks)

• Link-State Routing: A class of dynamic routing protocols (such as OSPF and IS-IS) where every router constructs a comprehensive topological map (Link-State Database / LSDB) of the network by flooding Link-State Advertisements (LSAs) and computing the shortest path using Dijkstra’s Shortest Path First (SPF) algorithm.
• OSPF LSA Taxonomy:
  • Type 1 (Router LSA): Generated by every internal router to describe its directly connected links within an area.
  • Type 2 (Network LSA): Generated by the Designated Router (DR) on multi-access broadcast segments.
  • Type 3 (Summary LSA): Generated by Area Border Routers (ABRs) to summarize routes between OSPF areas.
  • Type 4 (ASBR Summary LSA): Injected by ABRs to announce the path to an Autonomous System Boundary Router (ASBR).
  • Type 5 (AS External LSA): Injected by ASBRs to advertise routes redistributed from external autonomous systems.
  • Type 7 (NSSA External LSA): Injected within a Not-So-Stubby Area (NSSA) and converted to Type 5 by the ABR.
• Autonomous System (AS): A collection of IP networks and routers under a single administrative domain operating a common interior gateway routing policy (identified by a globally unique 16-bit or 32-bit ASN assigned by IANA/APNIC).
• Border Gateway Protocol (BGP-4): The de facto inter-domain path-vector routing protocol (RFC 4271) that exchanges Network Layer Reachability Information (NLRI) over reliable TCP connections (port 179) between independent Autonomous Systems.
• BGP Path Attributes:
  • Well-Known Mandatory: AS_PATH, NEXT_HOP, ORIGIN (must be present in every BGP UPDATE message).
  • Well-Known Discretionary: LOCAL_PREF, ATOMIC_AGGREGATE (recognized by all implementations; presence is optional).
  • Optional Transitive: COMMUNITY, AGGREGATOR (passed to downstream peers even if unrecognized).
  • Optional Non-Transitive: Multi-Exit Discriminator (MED / metric) (evaluated locally; stripped when crossing AS boundaries).

---

Part 2: Structural Frameworks (10 Marks)

1. Multi-Area OSPF vs. Autonomous System Inter-Domain Architecture

AUTONOMOUS SYSTEM (AS 100)

• +--------------------+ +--------------------+
• | Stub Area 1 | | NSSA Area 2 |
• | Internal Routers | | ASBR -> External |
• +---------+----------+ +---------+----------+
• | (Type 3 LSA) | (Type 7 LSA)
• v v
• +---------+-------------------------------+----------+
• | Backbone Area 0 (Transit Core) |
• | ABR 1 ABR 2 |
• +-------------------------+---------------------------+
• | (Edge eBGP Session)
• TCP Port 179
• v
• AS 200 (Upstream ISP / NPIX Peering Exchange)

2. Complete 8-Step BGP Best Path Selection Decision Process

When a BGP router receives multiple paths toward the same NLRI prefix, it executes the following deterministic elimination ladder:

1. Highest Weight: Cisco/vendor-proprietary local parameter (0-65535, default 32768 for locally generated routes).
2. Highest LOCAL_PREF: Standard attribute (default 100) evaluated across the entire iBGP domain to choose outbound exit point.
3. Locally Originated: Prefer routes originated locally via network or aggregate-address statements over redistributed routes.
4. Shortest AS_PATH: Path with the lowest number of AS hops (AS prepending increases length to discourage inbound traffic).
5. Lowest ORIGIN Code: Prefer IGP (i) over EGP (e), and EGP over Incomplete (?).
6. Lowest MED: Influences inbound traffic from neighboring AS (only compared if neighbor AS is identical).
7. eBGP over iBGP: Prefer external eBGP path over internal iBGP learned path.
8. Lowest IGP Metric to BGP NEXT_HOP: Path reachable via the closest internal IGP next hop.
9. Tie-Breakers: Lowest BGP Router ID, oldest eBGP route, and lowest neighbor IP address.

---

Part 3: Distinction Tables

---

Part 4: Worked Problems & Exam Traps

Problem Archetype: Calculating OSPF Cost and Resolving iBGP Full-Mesh Scaling

Question: An enterprise network has a 10 Gbps core link and a 1 Gbps branch link. Using default reference bandwidth ($100\text{ Mbps}$), what is the cost for each? How do you fix metric differentiation? Furthermore, if an enterprise has 12 internal routers running iBGP, how many peering sessions are required without Route Reflectors? Solution:

1. Default OSPF Cost Formula:

   Swipe to see the full formula↔

   $$\text{Cost} = \frac{\text{Reference Bandwidth}}{\text{Interface Bandwidth}} = \frac{10^8}{\text{Bandwidth}}$$
   • For 1 Gbps ($10^9$ bps): $\text{Cost} = \frac{10^8}{10^9} = 0.1 \rightarrow 1$ (minimum integer cost is 1).
   • For 10 Gbps ($10^{10}$ bps): $\text{Cost} = \frac{10^8}{10^{10}} = 0.01 \rightarrow 1$.
   • Problem: Both 1 Gbps and 10 Gbps links receive cost 1!
   • Remedy: Reconfigure reference bandwidth on all routers: auto-cost reference-bandwidth 100000 (100 Gbps), yielding Cost 100 for 1 Gbps and Cost 10 for 10 Gbps.
2. iBGP Full-Mesh Formula:

   Swipe to see the full formula↔

   $$\text{Sessions} = \frac{N(N-1)}{2} = \frac{12 \times 11}{2} = 66\text{ peering sessions}$$
   • Scaling Remedy: Deploy BGP Route Reflectors (RR) or BGP Confederations to reduce peering complexity from $O(N^2)$ to $O(N)$.

TU Exam Trap: In BGP exam questions, students frequently claim that BGP uses hop count like RIP because it uses AS_PATH. This is mathematically false: AS_PATH counts autonomous system boundaries traversed, not individual router hops. A route with 1 AS hop through 50 core routers is strictly preferred over 2 AS hops through 2 routers! Furthermore, remember that LOCAL_PREF exits your AS (outbound traffic control), while AS_PATH prepending influences external traffic entering your AS (inbound traffic control).