This paper explores the use of genetic algorithms (GAs) for designing networks with time-varying demands. The study investigates how different GA operators affect the quality of the best network found, focusing on the trade-off between pleiotropy (servers performing multiple tasks) and redundancy (multiple servers performing the same task). The quality of the network is measured by cost and reliability. The paper discusses the advantages of GAs over traditional optimization methods, including adaptability to changing problem constraints and the ability to explore a wide search space. It also highlights the importance of crossover operators in improving the search process, as they allow the algorithm to explore a wider range of networks compared to mutation operators alone. The study presents a network model with servers, clients, and links, and defines a fitness function based on reliability and cost. The fitness function is calculated using Dijkstra's algorithm to determine the shortest path between nodes. The paper also defines redundancy and pleiotropy measures to evaluate the network's properties. The results show that the crossover operator allows the GA to converge faster and explore a wider range of networks. The study also finds that the convergence time is not significantly affected by population size or link failure probability, although it tends to decrease with higher link reliability. The paper concludes that further research is needed to refine the convergence time and to optimize for reliability and cost separately, combining populations using the crossover operator. The study also suggests benchmarking the GA on small networks where the optimal solution is known.This paper explores the use of genetic algorithms (GAs) for designing networks with time-varying demands. The study investigates how different GA operators affect the quality of the best network found, focusing on the trade-off between pleiotropy (servers performing multiple tasks) and redundancy (multiple servers performing the same task). The quality of the network is measured by cost and reliability. The paper discusses the advantages of GAs over traditional optimization methods, including adaptability to changing problem constraints and the ability to explore a wide search space. It also highlights the importance of crossover operators in improving the search process, as they allow the algorithm to explore a wider range of networks compared to mutation operators alone. The study presents a network model with servers, clients, and links, and defines a fitness function based on reliability and cost. The fitness function is calculated using Dijkstra's algorithm to determine the shortest path between nodes. The paper also defines redundancy and pleiotropy measures to evaluate the network's properties. The results show that the crossover operator allows the GA to converge faster and explore a wider range of networks. The study also finds that the convergence time is not significantly affected by population size or link failure probability, although it tends to decrease with higher link reliability. The paper concludes that further research is needed to refine the convergence time and to optimize for reliability and cost separately, combining populations using the crossover operator. The study also suggests benchmarking the GA on small networks where the optimal solution is known.