Actively learning PoPs to probe and probing frequency to maximize application experience predictions

An apparatus, comprising:

• one or more network interfaces;
• a processor coupled to the one or more network interfaces and configured to execute one or more processes;
• and
• a memory configured to store a process that is executable by the processor, the process when executed configured to:
• compute, and for each of a set of points of presence via which traffic for an online application can be sent from a location, application experience metrics predicted for the online application over time;
• assign, for each of the set of points of presence, weights to different time periods, based on measures of uncertainty associated with the application experience metrics predicted for the online application over time;
• generate, based on the weights assigned to the different time periods for each of the set of points of presence, schedules for probing network paths connecting the location to the online application via the set of points of presence, wherein the schedules allocate probing across the different time periods and network paths, to maximize their corresponding weights, given one or more probe quotas specified via a user interface;
• and
• cause the network paths to be probed in accordance with their schedules, wherein results of this probing are used to select a particular point of presence from among the set of points of presence via which traffic for the online application should be sent from the location during a certain time period.

A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute a process comprising:

• computing, by the device and for each of a set of points of presence via which traffic for an online application can be sent from a location, application experience metrics predicted for the online application over time;
• assigning, by the device and for each of the set of points of presence, weights to different time periods, based on measures of uncertainty associated with the application experience metrics predicted for the online application over time;
• generating, by the device and based on the weights assigned to the different time periods for each of the set of points of presence, schedules for probing network paths connecting the location to the online application via the set of points of presence, wherein the schedules allocate probing across the different time periods and network paths, to maximize their corresponding weights, given one or more probe quotas specified via a user interface;
• and
• causing, by the device, the network paths to be probed in accordance with their schedules, wherein results of this probing are used to select a particular point of presence from among the set of points of presence via which traffic for the online application should be sent from the location during a certain time period.