Traffic Engineering

1
Traffic Engineering

• By
• Kavitha Ganapa

2
Introduction

• Traffic engineering is concerned with the issue
  of performance evaluation and optimization of
  operational IP networks
• Through carefully managing the traffic
  distribution inside a network
• Congestion hot spots can be reduced
• Resource utilization can be improved

3
Need for Traffic engineering

• Need for service providers to efficiently provide
  and manage network resources to accommodate
  resource commitments
• To meet customers expectation of guaranteed and
  differentiated services
• To reduce cost of provisioning the services

4
Fish problem

• Leads to extremely unbalanced traffic
• Poor utilization of network resources

5
Fish problem in IP routing

• Fish problem is caused by primarily two
  properties
• IP routing is destination-based, all packets
  whose destination addresses share the same prefix
  have the same next hop
• Decision making in current routing is based on
  local optimization

6
Traffic engineering solutions

• Fish problem is solved by going beyond current
  destination based routing and by providing
  mechanisms to explicitly manage the traffic
  inside network
• Overlay Model
• Peer Model

7
Overlay Model

• Calculates the routes for traffic demands for
  edge nodes
• Service providers then set up MPLS explicit
  routes between edge routes and match traffic
  demands over them, creating a full-mesh network
• Suffers from N-square problem, problems with
  scalability

8
Peer Model

• Achieves balanced traffic distribution by
  manipulating link weights in the OSPF routing
  protocol
• More scalable than Overlay Model
• Packets are forwarded based on longest-prefix
  match
• Eliminates N-square problem and reduces
• Reduces messaging overheads in setting up
  explicit routes

9
Optimization objectives

• Minimizing congestion and packet losses in the
  network
• Network congestion is caused by
• Inadequate network resources (can be solved by
  new capacity or reduce demands)
• Unbalanced traffic distribution (can be addressed
  by better management of the resources in the
  network)
• Improving link Utilization
• Minimizing the total delay experienced by packets
• Increasing the number of customers with the
  current assets

10
One objective

• To minimize the maximum of link utilization
• Why?
• Hot spots are the points with the highest link
  utilization
• Queuing delay increases nonlinearly as link
  utilization becomes higher
• By minimizing the maximum of link utilization
  traffic is spread and the resources are
  efficiently used

11
Advantages of the Objective

• Reduces the total delay experienced by the
  packets
• Moves the traffic away from congested hot spots
  to less utilized parts of the network thus
  balanced traffic distribution
• Leaves more space for future traffic growth, the
  percentage of residual bandwidth is maximized by
  the minimizing the maximum link utilization

12
Building blocks

• Data repository
• Topology and state discovery
• Traffic demand estimation
• Route computation (constraint-based routing)
• Network interface
• Graphic user interface
• Topology and state discovery and constraint based
  routing are two critical components of traffic
  engineering

13
Topology and state discovery

• OSPF-based scheme
• Because OSPF is widely deployed
• Has the necessary mechanisms for distributing
  link status and constructing a topology database
• Extended link state information for traffic
  engineering
• Local and remote interface IP addresses
• Traffic engineering metric
• Maximum bandwidth
• Maximum reservable bandwidth
• Unreserved bandwidth
• Resource class

14
Constraint-based routing

• Shortest path
• Minimum hop
• Minimize the total resource consumption per route
• Shortest widest path
• Avoids overloading by maximizing the residual
  capacity across the network
• Hybrid algorithm
• Combination of Shortest path/Minimum hop and
  shortest widest path

15
Multipath load sharing

• Hashing based traffic splitting schemes
• Direct hashing
• Can split a load only into equal amounts
• Table-based hashing
• Can distribute the traffic as required

16
Summary

• Traffic engineering is concerned with performance
  optimization of operational networks
• Provides a set of tools for efficiently
  provisioning and managing backbone networks
• Used to reduce congestion hot spots, improve
  overall utilization of networks and reduce the
  cost for meeting resource commitments to their
  customers

17
Research work

• Our work includes the investigation of the
  QoS-based routing problem under two different
  frameworks
• the traditionally used SPC-QRF
• The newly proposed SPD-QRF

18
Acknowledgments

• The slides are prepared based on the information
  from Architectures and Mechanisms for Quality of
  Service by Zheng Wang