Quality of Service (QoS) Architectures in the Modern Internet

Unit IV explores the two primary models for providing Quality of Service (QoS) in IP-based networks: Integrated Services (IntServ) and Differentiated Services (DiffServ). As high-speed networks evolved, the basic "Best Effort" delivery model of the Internet became insufficient for real-time applications like VoIP, telemedicine, and financial high-frequency trading.

Key Focus Areas:

1. Integrated Services (IntServ): The "Hard QoS" model based on per-flow resource reservation.
2. RSVP Signaling: How applications request network resources.
3. Queuing Disciplines: The mathematical and practical study of FQ, GPS, and WFQ.
4. Differentiated Services (DiffServ): The "Soft QoS" model based on aggregate traffic classes and per-hop behaviors.
5. Random Early Detection (RED): Advanced buffer management to prevent global synchronization.
6. Traffic Conditioning: Policing, shaping, and marking at the network edge.

---

4.0.1 The Evolution of QoS

• Best Effort: Original IP model. No guarantees. First-in-first-out (FIFO).
• IntServ (RFC 1633): Designed for end-to-end guarantees. Scales poorly because routers must track every individual flow.
• DiffServ (RFC 2475): Scalable alternative. Routers handle groups of traffic (e.g., all voice traffic) rather than individual sessions.

4.0.2 Unit IV Roadmap

This unit provides the algorithmic depth required to manage congested high-speed backbones:

• Lesson 4.1: IntServ components and RSVP signaling.
• Lesson 4.2: Theoretical queuing models (GPS and WFQ).
• Lesson 4.3: DiffServ architecture and Traffic Conditioners.
• Lesson 4.4: RED buffer management and mathematical derivations.