REPETITA

Groups topology data, traffic matrices, link capacities, delays, IGP weights, Segment Routing information, and explicit paths into complete problem instances that solvers and scenarios can process consistently during controlled traffic-engineering experiments across many network configurations and demand patterns.

Uses Gurobi to solve a mixed integer linear program that selects Segment Routing detours, minimizes maximum link utilization, limits each demand to one detour, assumes even load balancing, and reduces variables through an efficient modified node-link formulation.

Applies local search to identify link weights that minimize maximum utilization, supports shortest-path routing comparisons, and uses reusable traffic-engineering libraries for path computation, link-load calculation, and concise implementation of established weight-optimization methods directly within reproducible experimental settings.

Uses a randomized gravity model to synthesize realistic demand matrices, creates five matrices for every topology, and scales traffic so optimal maximum link utilization reaches ninety percent, producing challenging but adjustable test conditions for evaluating congestion-optimization algorithms.

Provides more than 250 realistic networks collected from Rocketfuel, DEFO, and Internet Topology Zoo sources, with missing capacities completed, extreme capacity variations normalized, disconnected components processed, and five consistently generated synthetic traffic matrices assigned to every supported topology.