UNIT III

Quality Of Service : Introduction - Quality of Service in IP Version 4 - Integrated Services - Differentiated

Services - Quality of Service with Nested Differentiated Services Levels - IP Multicast and Anycast:
Addressing - Multicast Routing - Routing Protocols – Anycasting- IPv6 Anycast Routing Protocol: Protocol
Independent Anycast—Sparse Mode - Transport over Packet: Draft-Martini Signaling and Encapsulation -
Layer-2 Tunneling Protocol.

IP Quality of Service (QoS) – Introduction

The Internet Protocol (IP) was originally designed for best-effort service, meaning all packets
were treated equally without prioritization. However, as network usage has evolved—especially
with the rise of real-time applications like voice, video conferencing, and streaming—there is
a growing need for service differentiation to ensure better performance for certain types of
traffic.

Why is QoS Needed?

1. Different Applications Have Different Requirements

o Real-time applications (VoIP, Video Calls) → Require low latency & minimal packet loss
o File transfers (FTP, cloud backups) → Can tolerate delays but need high bandwidth

2. Network Congestion Affects Performance

o When multiple applications share a network, congestion can cause delays, jitter, and
packet loss, leading to poor user experience.

3. Not All Packets are Equal

o Some packets (e.g., emergency calls, financial transactions) may need higher priority
than others (e.g., background downloads).

QoS Mechanisms in IP Networks

Several techniques have been proposed to handle traffic service differentiation in IP networks:

1. Best-Effort Service
o Traditional IP approach: No priority, first-come, first-served.
o Works well for general data traffic but not for real-time services.

2. Integrated Services (IntServ)

o Uses resource reservation to guarantee QoS for applications.
o Requires network-wide signaling (RSVP protocol), making it complex to scale.

3. Differentiated Services (DiffServ)

 o Assigns different priority levels to traffic classes.
o Uses DSCP (Differentiated Services Code Point) field in IP headers to classify packets.
o Scalable and widely used in modern networks.

4. Traffic Shaping & Policing

o Traffic shaping: Controls data transmission rate to prevent congestion.
o Policing: Drops or marks packets exceeding allowed bandwidth limits.

5. Queuing & Scheduling

o Priority Queuing (PQ): High-priority packets get processed first.
o Weighted Fair Queuing (WFQ): Ensures bandwidth fairness among flows.

Quality of Service (QoS) in IPv4

In IPv4, the Type of Service (TOS) field in the datagram header (8 bits) was originally designed
to specify how a packet should be handled by routers. However, over time, the Differentiated
Services (DiffServ) model replaced this TOS approach, refining traffic classification and
prioritization.

Original IPv4 Type of Service (TOS) Field (RFC 791, RFC 1349)

The 8-bit TOS field was structured as follows:

Precedence Levels (Bits 0–2)

  Higher precedence packets (e.g., Network Control) were given priority over lower
precedence ones (e.g., Routine).
 The remaining four bits (3-6) were used for additional service preferences.

Modern Differentiated Services (DiffServ) Model

 The TOS field has been replaced with the Differentiated Services Field (DS Field).
 The first 6 bits are called the Differentiated Services Code Point (DSCP).
 The last 2 bits are reserved for Explicit Congestion Notification (ECN).

DSCP Format

 The DSCP value determines how a packet is treated by routers and falls into different classes.
 The ECN bits allow routers to signal congestion without dropping packets.

QoS Classes in DiffServ

DiffServ defines multiple service classes:

1. Best Effort (BE) → Default, no special QoS treatment.

2. Assured Forwarding (AF) → Prioritizes packets into 12 classes (AF11-AF43).
3. Expedited Forwarding (EF) → Used for low-latency traffic (e.g., VoIP).
Example DSCP Values

Integrated Services (IntServ) – A QoS Paradigm

Integrated Services (IntServ) is a Quality of Service (QoS) model that provides explicit flow-
based resource allocation in IP networks. It ensures that specific traffic flows receive
guaranteed network performance, such as low latency, minimal jitter, and controlled
bandwidth.

Key Concepts of IntServ

1. Flows & Datagram Treatment

o Packets with similar characteristics (e.g., same source-destination pair) are grouped into
flows.
o TOS bits in IPv4 help determine the level of QoS assigned to these flows.

2. QoS Parameters Considered

o Bandwidth → Amount of data that can be transmitted per second.
o Delay (Latency) → Time taken for a packet to reach its destination.
o Jitter → Variation in packet delay (important for real-time applications).
o Reliability → Packet loss rate and error handling mechanisms.

3. Controlled Link Sharing

 o Network operators divide traffic into administrative classes (e.g., VoIP, Video
Streaming, General Browsing).
o A minimum percentage of bandwidth is assigned to each class during congestion.
o Unused bandwidth is dynamically allocated to other classes when available.

Quality of Service (QoS) architecture for an Integrated Services (IntServ) model in a

network.

It consists of two primary layers:

1. Network Layer

 Routing Agent & Routing Database → Handles packet forwarding and path selection.
 Reservation Setup Agent → Manages QoS reservations (likely using RSVP).
 Admission Control Agent → Decides if new traffic flows can be admitted based on
available resources.
 Traffic Control Database → Stores traffic policies, ensuring QoS adherence.
 Management Agent → Oversees overall QoS operations.

2. Data-Link Layer

 Input Interface → Receives incoming packets.

 Classifier → Identifies and categorizes traffic based on QoS needs.
 Internet Forwarder → Sends classified packets to the scheduler.
 Packet Scheduler → Determines transmission priority and allocates bandwidth.
 Output Interface → Transmits packets based on QoS scheduling.
Packet Scheduler in Integrated Services (IntServ)

The packet scheduler is responsible for managing packet forwarding based on QoS
requirements. It determines when and how packets are transmitted using a set of queues,
timers, and scheduling algorithms.

Packet Classifier in Integrated Services (IntServ)

The packet classifier is responsible for identifying and mapping incoming packets into
predefined traffic classes. This classification determines how packets are treated by the packet
scheduler and other QoS mechanisms.

Key Functions of a Packet Classifier

1. Flow Identification
o Determines which flow an incoming packet belongs to.
o Can be based on source/destination IP, port number, protocol, or other fields.

2. Class Mapping
o Assigns packets to QoS classes (e.g., voice, video, best-effort, critical data).
o Higher-priority classes receive better service (lower delay, guaranteed bandwidth).

3. Header Inspection
o Uses packet headers (IPv4, IPv6, MPLS, etc.) to extract relevant classification data.
o May also use deep packet inspection (DPI) if additional info is required.

4. Hierarchical Classification
o Classification may differ at different network points (edge routers vs. core routers).
o Example: Edge routers classify individual flows, while core routers aggregate them into
fewer classes.

Admission Control in Integrated Services (IntServ)

Admission control is a crucial mechanism in QoS-enabled networks that decides whether a

new traffic flow can be admitted without violating the QoS guarantees of existing flows. It
ensures that network resources are not overcommitted, thereby maintaining service quality.

Key Functions of Admission Control

1. QoS Feasibility Check

o Determines whether there is enough bandwidth, buffer space, and processing power to
support the new flow.
o Ensures existing QoS commitments remain unaffected.
 2. Decision Making (Accept/Reject)
o If resources are available, the request is accepted, and the flow is admitted.
o If resources are insufficient, the request is rejected to prevent congestion and
degradation of service.

3. Policy Enforcement
o Ensures compliance with network policies (e.g., priority for premium users).
o Prevents unauthorized users from reserving excessive bandwidth.

4. Resource Reservation
o Works alongside Resource Reservation Protocol (RSVP) to allocate required resources.

5. Traffic Accounting & Reporting

o Tracks resource usage for billing and capacity planning.

Resource Reservation Protocol (RSVP) Overview

RSVP (Resource Reservation Protocol) is a signaling protocol used in IP networks to reserve

resources for specific data flows to provide Quality of Service (QoS). Defined in RFC 2205, it
allows applications to request bandwidth, delay guarantees, or other QoS parameters from
network elements.

Key Characteristics of RSVP

1. Flow-Specific Resource Reservation

o RSVP creates and maintains flow-specific states in end hosts and routers along the data
path.
o Resources are allocated for simplex (one-way) flows.

2. Signaling Mechanism for QoS

o Applications specify QoS requirements via a flowspec, which consists of:
 Rspec (Reserve Spec) → Defines desired QoS.
 Tspec (Traffic Spec) → Describes traffic characteristics.
o These parameters are passed to admission control and packet schedulers.

3. Works with IPv4 & IPv6

o RSVP operates on top of IPv4/IPv6, similar to ICMP or IGMP, but does not carry
application data.

4. Supports Unicast & Multicast Routing

o RSVP dynamically adapts to changes in multicast group membership.
o Uses IGMP for multicast group membership management.

5. Soft-State Mechanism
o RSVP periodically refreshes reservations.
o If refresh messages are not received, reservations automatically time out.
RSVP Message Types

Path Message Process:

 Sent from sender to receiver along routing paths.

 Stores path state at each router.

🔹 Resv Message Process:

 Sent from receiver to sender on the reverse path.

 Establishes resource reservations at each router.
Differentiated Services (DiffServ) Overview

Differentiated Services (DiffServ) is a scalable Quality of Service (QoS) model designed to

handle high-speed networks by aggregating multiple flows into a few traffic classes. Unlike
Integrated Services (IntServ), which requires per-flow reservations (via RSVP), DiffServ
aggregates traffic based on class-based rules, significantly reducing complexity.

Key Concepts of DiffServ

1. Traffic Classification & Aggregation

 DiffServ assigns packets to Behavior Aggregates (BAs) based on traffic classification.

 Each packet is marked with a Differentiated Services Code Point (DSCP) in the DS field of the
IPv4 TOS field (or IPv6 DS field).
  DiffServ-capable routers forward packets based on Per-Hop Behavior (PHB) associated with the
DSCP value.

2. Per-Hop Behavior (PHB) Types

3. Traffic Conditioning in DiffServ

To ensure traffic adheres to network policies, DiffServ applies conditioning mechanisms such
as:

Quality of Service with Nested Differentiated Services Levels

1. The Challenge: IntServ vs. DiffServ

 Integrated Services (IntServ) provides per-flow QoS, ensuring precise service guarantees but is
not scalable due to high resource demands.
 Differentiated Services (DiffServ) offers scalability by aggregating flows into service classes but
lacks per-flow QoS guarantees.
 A hybrid approach, IntServ over DiffServ, aims to map individual flow QoS requirements onto
aggregated DiffServ classes, but this reduces QoS granularity.
2. Nested DiffServ Model & Service Vectors (SV)

 Nested DiffServ introduces sub-levels within each DiffServ class to group flows with similar QoS
needs.
 Explicit Endpoint Admission Control (EEAC) uses Service Vectors (SVs) to dynamically adjust
QoS allocation per flow.
 Each router assigns a different service level (si, sj, etc.) for a flow, creating a Service Vector (SV)
= (s0, s1, ..., sm) that optimizes utility vs. cost.

3. Two-Phase EEAC Process

1. Probing Phase
o The source sends probe packets to the destination to collect service state of
intermediate routers.
o The destination host sends feedback, allowing the source to compute best QoS class
assignment at each router.
2. Data Transmission Phase
o The service vector (SV) is marked in each packet.
o Each router follows the assigned QoS level to balance cost and network utilization.

4. Impact on QoS Granularity

 QoS granularity is improved to O(p^q), where:

o p = Number of routers.
o q = Number of service classes.
 This reduces cost for users while optimizing network resource utilization for providers.

5. Integrating OSPF with Service Vectors for QoS Routing

 Traditional Open Shortest Path First (OSPF) can lead to suboptimal routing and inefficient
network utilization.
 Enhanced OSPF + Service Vectors (OSPF-SV) allows routers to:
o Detect available service classes for each link.
o Disseminate QoS-aware link state information.
o Improve path selection, balancing service quality and cost efficiency

Drawbacks of Explicit Endpoint Admission Control (EEAC) with Path Selection

The Explicit Endpoint Admission Control (EEAC) scheme relies on probing to determine
service availability along a selected path. However, when used with Open Shortest Path First
(OSPF) for routing, several limitations arise, leading to suboptimal QoS performance and
inefficient network resource utilization.
 1. Inaccurate Link Weight Assignment in OSPF

 Many vendors set OSPF link weights inversely proportional to link capacity, which does not
reflect real-time QoS conditions (e.g., delay, jitter, or packet loss).
 Some studies suggest mapping weights to QoS parameters (delay, bandwidth), but:
o There is no standard weight-setting formula.
o QoS is multi-dimensional, making weight computation complex.

Impact on EEAC

 OSPF selects paths based only on weights, ignoring actual QoS service class availability.
 EEAC probes the OSPF-selected path, but if no service vector (SV) meets the QoS requirement,
the request is denied—even when an alternative path exists.

2. False Routing Decisions Leading to Rejected Requests

Example 1: Flow from N1 to N6 (Figure 6.4)

Scenario:

 A flow requires delay < 25ms.

 OSPF selects the shortest path (P1 = N1 → N2 → N6) based on its weight-based cost calculation.
 EEAC probes P1 but finds no valid SV that meets the QoS requirement.
 Request is denied, even though other paths (P2, P3) could satisfy the requirement.

Better Alternative Paths Exist:

 P2 = (N1 → N5 → N6)
o Service vector (S2, S2), cost = 6 (optimal solution)
 P3 = (N1 → N3 → N5 → N6)
o Higher cost but still meets the QoS requirement.

🔴 Problem:
OSPF failed to select P2, leading to an unnecessary request denial.
Suboptimal Path Selection and Higher Costs

Example 2: Using Delay as Link Weight

 Suppose OSPF weights are adjusted to delay-based values.

 OSPF now selects P4 = (N1 → N3 → N4 → N6) as the shortest path.
 EEAC selects SV (S1, S2, S2) or (S2, S1, S2), with cost = 12.
 However, the optimal path remains P2, with cost = 6.

🔴 Problem:

 Suboptimal routing increases costs.

 Reduces network utilization efficiency.

OSPF-Based Adaptive and Flexible QoS Provisioning

To address the drawbacks of Explicit Endpoint Admission Control (EEAC) with traditional
OSPF, an improved architecture integrates OSPF with Service Vectors (SVs). This approach
enables more adaptive and QoS-aware routing.

. Enhanced Link Weight Representation

Instead of using a single scalar value to represent OSPF link weights, the proposed model
introduces:

 A vector-based weight representation, where each link documents all available service classes.
  A service class is defined as:

Benefits:

 Links with multiple service levels enable routers to make better routing decisions.
 Reduces unnecessary call denials by allowing lower service classes when higher ones
are unavailable.
 Service classes can be compared and sorted to facilitate efficient path selection.

2. Improved Path Selection Strategy

 The edge router (source node) selects a shortest path that satisfies user QoS constraints.
 Unlike generic QoS routing, the edge router also selects the service classes for the path.
 The selected SV (service vector) is marked in data packets for routers to provide appropriate
QoS.
 If no SV satisfies the user request, the request is denied.

3. Efficient Link-State Updates & Class-Based Triggering

Traditional OSPF Link Updates

 Fixed update interval (30 minutes)

 Too long for high-speed networks, leading to outdated QoS information.

Proposed Class-Based Triggering Mechanism

 Updates occur only when a link’s state crosses a service level boundary.
 Reduces excessive update overhead while maintaining accurate QoS information.

IP Multicast and Anycast: Enhancing Network Efficiency

Most internet traffic follows a unicast model, where data is sent from one source to one
destination. However, more efficient communication models exist for specific use cases,
including multicast and anycast.

1. IP Multicast: One-to-Many Communication

Definition:
Multicasting allows one source device to send data to multiple receivers simultaneously
without duplicating traffic for each recipient. This is more efficient than unicast or broadcast.

Addressing

Multicast Addressing in IPv4 and IPv6

1. IPv4 Multicast Addressing

In IPv4, Class D addresses are reserved for multicast communication. These addresses:

 Begin with the binary prefix 1110 (first octet: 224–239).

 Range from 224.0.0.0 to 239.255.255.255.
 Are only used as destination addresses, never as a source.

IPv4 Multicast Address Subdivisions

Multicast addresses are divided into different blocks based on their scope and function:

 224.0.0.0 – 224.0.0.255 → Reserved for local network control (e.g., OSPF, RIP).
 224.0.1.0 – 238.255.255.255 → Used for global multicast applications (e.g., streaming).
 239.0.0.0 – 239.255.255.255 → Administratively scoped addresses (private multicast).

🔹 Example: 224.0.0.1 is used for Network Time Protocol (NTP) synchronization.

2. IPv6 Multicast Addressing

In IPv6, multicast addresses start with 11111111 (FF in hexadecimal), reserving 1/256 of the
entire address space for multicast.

🔹 IPv6 Multicast Format:

FFXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX:XXXX

Where:

 FF → Identifies the address as multicast.

 XX → Defines the scope (e.g., link-local, site-local, global).
 Remaining fields → Used for multicast group identification.

🔹 Example: FF02::1 is used for all nodes on the local link.

Differences between Multicasting and Multiple Unicasting

Multicast Routing: Shortest-Path Trees (SPT)

Multicast routing is different from unicast routing because a multicast packet is intended for
multiple receivers rather than a single destination. The goal is to deliver packets efficiently
using a Shortest-Path Tree (SPT).
Key Concepts:

1. Shortest-Path Tree (SPT)

o The source node is the root of the tree.
o The leaves are the destination nodes (multicast receivers).
o The shortest path from the source to each receiver is determined using routing
protocols (e.g., OSPF, Dijkstra's algorithm).

Unicast Routing

Unicast routing is the process of sending a packet from a single source to a single destination
across a network. It is the most common form of routing used in the Internet.

Unicast Routing Process

1. A source sends a packet with the destination IP address.

2. The router checks its routing table to determine the next hop.
3. The packet is forwarded step-by-step until it reaches the destination.

Multicast Routing

Multicast routing is the process of sending a single packet from a source to multiple
destinations that are part of a multicast group. This allows for efficient bandwidth usage since
the same data is only transmitted once and then replicated by routers as needed.

Multicast Routing Approaches

1. Source-Based Tree Approach (Shortest-Path Trees - SPT)

 Each multicast group has a dedicated shortest-path tree (SPT) from the source to all receivers.
 Uses unicast routing information to construct the shortest path.
 Requires n shortest-path trees for n multicast groups, which can be complex.

2. Shared-Tree Approach

 A single multicast tree is shared across all multicast groups.

 Instead of having a separate SPT for each group, multicast data flows through a predefined root
or Rendezvous Point (RP).
 Reduces complexity but may lead to suboptimal paths.

Source-Based Tree (SPT) in Multicast Routing

The Source-Based Tree (SPT) approach constructs a dedicated shortest-path tree (SPT) for
each multicast group. This ensures that multicast packets travel along the shortest possible
paths from the source to all group members, optimizing bandwidth usage and minimizing
latency.
How Source-Based Trees Work

1. Each multicast group has its own SPT

o The router calculates a separate shortest-path tree for each multicast group.
o This is based on unicast routing information (e.g., OSPF, RIP, or BGP).
2. SPT Construction at Each Router
o If a network has n multicast groups, each router computes m SPTs (where m is
the number of groups it needs to handle).
o The routing table maintains a next-hop list for each multicast group.
3. Packet Forwarding Using SPT
o When a multicast packet arrives, the router looks up its SPT and forwards the
packet only to interfaces that lead to group members.