Dynamic network service data routing

A method, comprising:

• periodically measuring active network performance data for a network including a plurality of entry points to access a network service;
• calculate, by each of a plurality of experts, selection probabilities for each of the plurality of entry points, based, in part, on the measured active network performance data;
• select, by a reinforcement learning algorithm, an expected best entry point from the plurality of entry points, based on the selection probabilities, wherein selecting the expected best entry point comprises weighting the plurality of experts based in part on previously measured network performance data and selecting the expected best entry point based on the weighted outputs of the plurality of experts;
• route data associated with the network service to the selected entry point;
• measure passive network performance data for the selected entry point;
• and
• update the reinforcement learning algorithm, based in part on the measured passive network performance data.

A non-transitory, computer-readable medium including instructions that, when executed by processing circuitry, cause a device to:

• periodically transmit network probes to a plurality of entry points that access a network service;
• calculate, by each of a plurality of regression models, expected rewards for the plurality of entry points, based in part on measured network performance data from the network probes;
• determine a probability distribution of the plurality of regression models, the probability distribution representing the likelihood of selecting each entry point of the plurality of entry points as the best entry point;
• select an expected best entry point from the plurality of entry points, based on the probability distribution;
• route data associated with the network service to the selected entry point;
• measure passive network performance data for the selected entry point;
• and
• update the reinforcement learning algorithm, based in part on the measured passive network performance data.