Routing Planning for All-Optical Networks with Computing-Network Integration: Research on Low-Latency Path Optimization Using Improved Genetic Algorithms

In a computing-network integrated environment, all-optical networks must simultaneously meet the requirements of computing power collaboration, bandwidth allocation, and end-to-end low latency. However, under conditions of multi-service concurrency and multiple constraints, problems such as excessive path selection latency, low algorithm convergence efficiency, and uneven resource utilization can easily arise. To address these challenges, this paper unifies the abstraction of computing nodes and optical network topology, constructs a joint optimization model with minimum end-to-end latency as the core objective, and considers both link bandwidth and computing power reachability. An improved Genetic Algorithm (GA) low-latency routing planning method is proposed. Methodologically, a computing-network-aware path encoding method is used to complete individual representations. A minimum latency guidance strategy is introduced in the initialization phase to improve the quality of the initial solution. The search range is dynamically adjusted through adaptive crossover and mutation operators, and a combination of elite retention and local search mechanisms achieves synergy between global exploration and local optimization, ultimately completing the joint selection of optical paths and computing nodes. Simulation results show that, under typical all-optical network topologies, this method reduces end-to-end latency by 22.3% compared to SP and 12.9% compared to CA-SP. The highest service blocking rate is only 10.1%, and the average number of convergence generations is reduced by about 25.9% compared to CA-SP. This verifies its effectiveness in low-latency routing planning in integrated computing and networking all-optical networks.