A Traffic Engineering Attribute for BGP

Open Access
Arnold, Todd
Graduate Program:
Computer Science and Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 03, 2008
Committee Members:
  • George Kesidis, Thesis Advisor
  • BGP
  • routing
  • networking
  • MPLS
In the Internet, ``traffic engineering' seeks to satisfy a packet stream's, or flow's, quality of service requirements while simultaneously leveraging existing resources to distribute a network's traffic load across all available paths, in order to provide congestion control. Traditional interior gateway protocols were expanded to distribute detailed interface attributes which are used by a constrained shortest path first algorithm to calculate a path which will best satisfy a flow's quality requirements. Multi-protocol label switching and reservation protocols are used to allocate resources along the path and to ensure the flow will traverse the calculated path. Traffic engineering is also performed between domains using the interdomain routing protocol, BGP. However, BGP's ability to balance network requirements across multiple links is severely limited. BGP's limited abilities only provide the means for designating the desired ingress point for a destination network. To rectify interdomain traffic engineering shortfalls, path reservations for MPLS using RSVP-TE was extended to cross domain boundaries. Despite the ability to reserve resources across domain boundaries, a method for distributing information to calculate which path will best satisfy a flow's QoS requirements across domain boundaries does not exist. We present a BGP attribute for distributing path quality information across domain boundaries. We explore a method for calculating the values to advertise while preserving the requirement to abstract a domain's internal policies. We also investigate the impact on routing table memory requirements for distributing this information by using simulations based on real-world topology and prefix information.