Abstract

Inter-AS link failures in conventional routing architectures are recovered through a combination of Border Gateway Protocol (BGP) hold-timer expiry, Link State Advertisement (LSA) flooding, and network-wide Forwarding Information Base (FIB) recomputation, a process that produces data plane outages measurable in seconds and affects every routing node in the involved Autonomous Systems (AS) regardless of topological relevance. This report presents the design, implementation, and evaluation of a failure recovery mechanism for Expedited Internet Bypass Protocol (EIBP), a label-based routing protocol in which forwarding decisions are derived from structured tier addresses rather than distributed routing tables. The implemented mechanism comprises three components: dual-path failure detection, a cached route deletion notification that propagates failure information from the failing Border Router through the Tier 1 node to every Tier 3 node in the affected AS, and a backup path discovery chain that allows Access Nodes to reroute traffic through a transit AS automatically on the first post-invalidation packet, without operator intervention. The implementation is evaluated against an Open Shortest Path First (OSPF)+BGP deployment on an identical three-AS ring topology hosted on the FABRIC virtualized testbed. Six failure scenarios are tested across three independent runs each, varying the direction of traffic flow and the identity of the failed Border Router. EIBP achieves a mean data plane convergence time of 0.591 seconds compared to 5.303 seconds for OSPF+BGP, a ninefold reduction. Control overhead is reduced by approximately 92 percent (1,126 bytes versus 14,411 bytes on average), and the recovery process affects 8 routing nodes compared to 14 for OSPF+BGP. The results demonstrate that EIBP’s targeted cached route deletion notification produces substantially faster and lower-cost failure recovery than the network-wide convergence process of the OSPF+BGP hybrid.

Publication Date

4-28-2026

Document Type

Master's Project

Student Type

Graduate

Degree Name

Information Technology and Analytics (MS)

Department, Program, or Center

Information Sciences and Technologies

College

Golisano College of Computing and Information Sciences

Advisor

Nirmala Shenoy

Advisor/Committee Member

Peter Willis

Campus

RIT – Main Campus

Download

Share

COinS