This paper explores the optimization of OSPF (Open Shortest Path First) weights to improve traffic engineering in internet routing. OSPF is the most commonly used intra-domain routing protocol, and it routes traffic along shortest paths based on link weights. The weights can be adjusted by network operators to avoid congestion, with a common heuristic being to set weights inversely proportional to link capacities. The authors analyze a proposed AT&T WorldNet backbone and show that optimizing OSPF weights for a given demand matrix is NP-hard, so they use a local search heuristic. Surprisingly, they find that the resulting OSPF routing performs within a few percent of the optimal general routing, which distributes traffic optimally across all paths. This challenges the common belief that OSPF leads to congestion and shows that for the studied network and demand matrix, switching to more flexible technologies like MPLS does not significantly improve load balancing. The study also tests the approach on synthetic internetworks based on a model by Zegura et al. While not always as close to the optimal solution, the local search heuristic outperforms standard heuristics like inverse capacity weighting or physical distance weighting, allowing for a 50%–110% increase in supported demand. The authors propose a local search heuristic that uses hashing tables to avoid cycling and improve search diversity. This approach is efficient and allows for rapid evaluation of OSPF routing costs. The heuristic is applied to both real and synthetic networks, demonstrating its effectiveness in improving traffic engineering performance. The results show that a clever OSPF weight setting can significantly enhance network performance, allowing it to handle up to 50%–110% more demand than traditional OSPF schemes. While MPLS offers more flexibility, the study suggests that OSPF with optimized weights can achieve substantial traffic engineering gains, especially for the studied networks and demand scenarios. The findings highlight the importance of traffic engineering in managing network resources efficiently and meeting service level agreements (SLAs) with customers.This paper explores the optimization of OSPF (Open Shortest Path First) weights to improve traffic engineering in internet routing. OSPF is the most commonly used intra-domain routing protocol, and it routes traffic along shortest paths based on link weights. The weights can be adjusted by network operators to avoid congestion, with a common heuristic being to set weights inversely proportional to link capacities. The authors analyze a proposed AT&T WorldNet backbone and show that optimizing OSPF weights for a given demand matrix is NP-hard, so they use a local search heuristic. Surprisingly, they find that the resulting OSPF routing performs within a few percent of the optimal general routing, which distributes traffic optimally across all paths. This challenges the common belief that OSPF leads to congestion and shows that for the studied network and demand matrix, switching to more flexible technologies like MPLS does not significantly improve load balancing. The study also tests the approach on synthetic internetworks based on a model by Zegura et al. While not always as close to the optimal solution, the local search heuristic outperforms standard heuristics like inverse capacity weighting or physical distance weighting, allowing for a 50%–110% increase in supported demand. The authors propose a local search heuristic that uses hashing tables to avoid cycling and improve search diversity. This approach is efficient and allows for rapid evaluation of OSPF routing costs. The heuristic is applied to both real and synthetic networks, demonstrating its effectiveness in improving traffic engineering performance. The results show that a clever OSPF weight setting can significantly enhance network performance, allowing it to handle up to 50%–110% more demand than traditional OSPF schemes. While MPLS offers more flexibility, the study suggests that OSPF with optimized weights can achieve substantial traffic engineering gains, especially for the studied networks and demand scenarios. The findings highlight the importance of traffic engineering in managing network resources efficiently and meeting service level agreements (SLAs) with customers.