Detailed Notes - Network Layer (Chapters 4 & 5)
=============================================
CHAPTER 4 - NETWORK LAYER: DATA PLANE
=============================================
1. Overview
The Network Layer provides logical communication between hosts across multiple networks. It hides
physical differences and offers services such as forwarding, routing, addressing, and packet
handling.
2. Data Plane vs Control Plane
* Data Plane: Handles per-packet forwarding. Implemented in hardware (ASICs) for high speed.
* Control Plane: Determines routes and populates forwarding tables. Implemented in software.
3. Router Architecture
Routers consist of:
- Input Ports: Receive and process incoming packets, perform header parsing and lookup.
- Switch Fabric: High-speed internal network transferring packets between input and output.
- Output Ports: Store and schedule packets for transmission.
- Routing Processor: Runs control-plane algorithms and management.
4. Forwarding
Forwarding tables map destination addresses to output ports.
Types:
* Destination-Based: Uses destination IP.
* Generalized Forwarding: Matches on multiple fields (IP, MAC, port, etc.) and applies actions.
5. Longest Prefix Matching (LPM)
Routers choose the route with the longest matching prefix to the destination IP.
Implemented using Trie structures or TCAMs for high-speed lookup.
6. Switching Fabrics
Types:
1. Switching via Memory (CPU-based, slow)
2. Switching via Bus (shared bus; limited by contention)
3. Switching via Interconnection Network (crossbar; high-performance)
7. Buffering and Queueing
When output links are busy, packets are queued.
* Input Buffering -> may cause Head-of-Line (HOL) blocking.
* Output Buffering -> multiple inputs contend for same output.
Queue Management: Tail Drop, RED, ECN.
Scheduling: FIFO, Priority, Round Robin, Weighted Fair Queuing (WFQ).
8. IP Packet Format & Fragmentation
IPv4 supports fragmentation; IPv6 removes it for routers.
IPv6 simplifies header (no checksum) and moves fragmentation to endpoints.
9. Addressing & DHCP
Dynamic Host Configuration Protocol automates IP allocation using:
Discover -> Offer -> Request -> ACK.
Provides address, gateway, DNS, subnet mask, and lease time.
10. NAT (Network Address Translation)
Maps private to public addresses using port translation.
Advantages: Conserves IPv4, improves security.
Disadvantages: Breaks end-to-end connectivity, complicates peer-to-peer apps.
11. IPv6
128-bit addressing, simplified header, no router fragmentation, faster forwarding.
Supports auto-configuration and dual-stack or tunneling transitions.
=============================================
CHAPTER 5 - NETWORK LAYER: CONTROL PLANE
=============================================
1. Overview
Control Plane determines packet paths across the network. It can be:
* Per-Router (distributed) - e.g., OSPF, BGP.
* Centralized (SDN) - controller computes and installs rules.
2. Routing Problem
Model network as a weighted graph.
Goal: Find least-cost paths for all source-destination pairs.
3. Link-State Routing (Dijkstra Algorithm)
Each router floods link-state info and independently runs Dijkstra's algorithm.
Pros: Fast convergence, loop-free.
Cons: Flooding overhead, more computation.
4. Distance-Vector Routing (Bellman-Ford Algorithm)
Routers exchange distance vectors with neighbors.
Equation: D(x, y) = min_v [ c(x, v) + D(v, y) ]
Pros: Simple.
Cons: Count-to-infinity problem, slow convergence.
5. Hierarchical Routing (OSPF)
OSPF divides network into areas with a backbone. Each area runs its own link-state algorithm.
Supports equal-cost multi-path and authentication.
6. Inter-Domain Routing (BGP)
BGP connects Autonomous Systems (AS). It is policy-driven (not purely shortest-path).
Attributes: AS-PATH, NEXT-HOP, LOCAL-PREF, MED.
Message Types: OPEN, UPDATE, KEEPALIVE, NOTIFICATION.
Route selection prioritizes local preference, AS path, and policy.
7. Software Defined Networking (SDN)
Separates control and data planes.
Architecture:
* Data Plane - simple forwarding devices.
* Control Plane - centralized controller.
* Applications - define policies.
OpenFlow protocol manages "match + action" flow tables.
Advantages: Programmability, flexibility.
Challenges: Scalability, reliability.
8. ICMP (Internet Control Message Protocol)
Used for error and control messages (e.g., ping, traceroute).
Common messages: Echo Request/Reply, Destination Unreachable, TTL Exceeded.
9. Network Management
SNMP: Queries device info via MIB, uses UDP (port 161).
NETCONF + YANG: XML/JSON-based management with transactional config changes.
10. Middleboxes
Perform functions beyond IP forwarding: NAT, firewall, load balancer, IDS.
Network Function Virtualization (NFV) enables software-based implementations.
11. Internet Architecture Principles
Internet's hourglass model has IP as the universal layer.
End-to-End Principle: place functionality at edges unless necessary in network core.
Middleboxes violate it but serve practical roles.
---------------------------------------------
Key Takeaways
---------------------------------------------
* Forwarding = Data Plane = Per-packet actions.
* Routing = Control Plane = Path computation.
* Dijkstra (LS) and Bellman-Ford (DV) are core routing algorithms.
* OSPF (intra-domain) vs BGP (inter-domain).
* SDN = centralized programmable networks.
* NAT, DHCP, IPv6 handle addressing and transition.
* ICMP, SNMP, NETCONF enable management and diagnostics.