US Patent 10193811 Flow distribution using telemetry and machine learning techniques

A device may receive, from a set of network devices, telemetry information associated with a first set of flows. The device may process, using a machine learning technique, the telemetry information to determine information that permits the set of network devices to identify a set of expected bandwidth values associated with a second set of flows. The device may provide, to a network device of the set of network devices, at least a portion of the information, that permits the set of network devices to identify the set of expected bandwidth values associated with the second set of flows, to permit the network device to determine an expected bandwidth value associated with a flow of the second set of flows and select a link to use when providing network traffic associated with the flow based on the expected bandwidth value.