SWAN (Software-driven WAN) is a system designed to enhance the utilization of inter-datacenter (DC) networks by centrally controlling traffic flow and frequently reconfiguring the network's data plane to match current demand. The system addresses the inefficiencies caused by uncoordinated service traffic and the distributed resource allocation model, which often results in underutilization of network capacity. SWAN introduces a novel technique that uses a small amount of scratch capacity on links to apply updates without causing transient congestion, ensuring that updates can be performed in a congestion-free manner. Additionally, SWAN dynamically selects a small set of forwarding rules to scale to large networks, leveraging limited forwarding table capacity. Experiments using a testbed prototype and data-driven simulations of two production networks show that SWAN can carry 60% more traffic than current practices, achieving near-optimal throughput and fair resource allocation. The system also enables congestion-controlled updates, ensuring that network updates do not disrupt traffic flow.SWAN (Software-driven WAN) is a system designed to enhance the utilization of inter-datacenter (DC) networks by centrally controlling traffic flow and frequently reconfiguring the network's data plane to match current demand. The system addresses the inefficiencies caused by uncoordinated service traffic and the distributed resource allocation model, which often results in underutilization of network capacity. SWAN introduces a novel technique that uses a small amount of scratch capacity on links to apply updates without causing transient congestion, ensuring that updates can be performed in a congestion-free manner. Additionally, SWAN dynamically selects a small set of forwarding rules to scale to large networks, leveraging limited forwarding table capacity. Experiments using a testbed prototype and data-driven simulations of two production networks show that SWAN can carry 60% more traffic than current practices, achieving near-optimal throughput and fair resource allocation. The system also enables congestion-controlled updates, ensuring that network updates do not disrupt traffic flow.