Performance Evaluation of Segment Routing Protocols in Modern Networks

Abstract

Next-generation networks require more efficiency in scaling and programming capabilities. Since SR facilitates effective path computation, service chaining, and traffic engineering, it appears essential for next-generation network routing. In this document, I introduce a novel evaluation framework called Segment Routing-Performance Evaluation and Routing Framework (SR-PERF) which aims to evaluate the operational effectiveness, scalability as well as Quality of Service (QoS) compliance of segment routing protocols-SR-MPLS and SRv6-in contemporary IP and SDN contexts. The SR-PERF model merges a multi-metric evaluation engine with key performance indicators such as path setup latency, end-to-end delay, control plane expenses including overhead, segment loss under dynamic links, and packet loss during dynamic link conditions. Using custom topologies within Mininet together with ONOS and FRRouting controllers enables simulation. Results presented show SRv6 improving on metrics such as backbone convergence time by 13.7% when compared to SR-MPLS alongside better adaptability during link drop scenarios despite slightly higher memory consumption due to header overhead stemming from IPv6. Moreover hybrid deployments incorporating intent-based path selection outperformed static OSPF/MPLGS baselines demonstrating improved throughput by 21% while reducing packet drops by 17%. This paper also studies the impact of increasing node density alongside traffic flow on the scalability of segment routing. The results point out that the control plane stability and the segment stack depth have a greater impact on SRv6 performance past 128 nodes, revealing some design compromises for hardware-limited systems. Lastly, described in this work are dashboards provided by the proposed SR-PERF framework which visualize routes and segments in real time aiding proactive segmentation optimization in programmable networks. The results also highlight that with SR-PERF it is possible to benchmark SR variants in a diverse setting enabling better informed selection and tuning of protocols for deployment within carrier-grade and cloud-native infrastructures.