Adaptive selection of network paths based on long-term predictions of user experience
In one embodiment, a device predicts, for each of a set of paths via which traffic for an online application can be routed, a distribution of an application experience metric for the online application. The device computes, for different subsets of the set of paths, aggregated distributions of their distributions of the application experience metric predicted by the device. The device makes comparisons between the aggregated distributions for the different subsets of the set of paths. The device causes, based on the comparisons, the traffic for the online application to be routed via a particular subset of the set of paths.
1. A method comprising:
predicting, by a device and for each of a set of paths via which traffic for an online application can be routed, a distribution of an application experience metric for the online application;
computing, by the device and for different subsets of the set of paths, aggregated distributions of their distributions of the application experience metric predicted by the device;
making, by the device, comparisons between the aggregated distributions for the different subsets of the set of paths; and
causing, by the device and based on the comparisons, the traffic for the online application to be routed via a particular subset of the set of paths.
2. The method as in claim 1 , wherein the application experience metric comprises at least one of: a transaction time for the online application, a mean opinion score (MOS) for the online application, or a user-specified satisfaction rating for the online application.
3. The method as in claim 1 , wherein the application experience metric is computed based on a probability of a service level agreement (SLA) for the online application being violated.
4. The method as in claim 1 , wherein causing the traffic for the online application to be routed via one of the different subsets of the set of paths comprises:
providing a recommendation to a user interface that the traffic of the online application should be routed via the particular subset.
5. The method as in claim 1 , wherein causing the traffic for the online application to be routed via the particular subset of the set of paths comprises:
causing the traffic for the online application to be rerouted from a default subset of the set of paths to the particular subset of the set of paths.
6. The method as in claim 5 , wherein making comparisons between the aggregated distributions for the different subsets of the set of paths comprises:
determining whether the traffic for the online application should be rerouted based on a distance between an aggregated distribution for the default subset and an aggregated distribution of the particular subset.
7. The method as in claim 5 , wherein making comparisons between the aggregated distributions for the different subsets of the set of paths comprises:
determining whether rerouting the traffic for the online application from the default subset of the set of paths to the particular subset of the set of paths would result in an improvement to the application experience metric for the online application that exceeds a threshold defined by a user-specified policy.
8. The method as in claim 1 , wherein paths in the particular subset are associated with a common site.
9. The method as in claim 1 , wherein the device makes comparisons between the aggregated distributions for the different subsets of the set of paths based in part on predicted traffic loads for the different subsets.
10. The method as in claim 1 , wherein the online application is a software-as-a-service (SaaS) application.
11. 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:
predict, for each of a set of paths via which traffic for an online application can be routed, a distribution of an application experience metric for the online application;
compute, for different subsets of the set of paths, aggregated distributions of their distributions of the application experience metric predicted by the apparatus;
make comparisons between the aggregated distributions for the different subsets of the set of paths; and
cause, based on the comparisons, the traffic for the online application to be routed via a particular subset of the set of paths.
12. The apparatus as in claim 11 , wherein the application experience metric comprises at least one of: a transaction time for the online application, a mean opinion score (MOS) for the online application, or a user-specified satisfaction rating for the online application.
13. The apparatus as in claim 11 , wherein the application experience metric is computed based on a probability of a service level agreement (SLA) for the online application being violated.
14. The apparatus as in claim 11 , wherein the apparatus causes the traffic for the online application to be routed via one of the different subsets of the set of paths by:
providing a recommendation to a user interface that the traffic of the online application should be routed via the particular subset.
15. The apparatus as in claim 11 , wherein the apparatus causes the traffic for the online application to be routed via the particular subset of the set of paths by:
causing the traffic for the online application to be rerouted from a default subset of the set of paths to the particular subset of the set of paths.
16. The apparatus as in claim 15 , wherein the apparatus makes comparisons between the aggregated distributions for the different subsets of the set of paths by:
determining whether the traffic for the online application should be rerouted based on a distance between an aggregated distribution for the default subset and an aggregated distribution of the particular subset.
17. The apparatus as in claim 15 , wherein the apparatus makes comparisons between the aggregated distributions for the different subsets of the set of paths by:
determining whether rerouting the traffic for the online application from the default subset of the set of paths to the particular subset of the set of paths would result in an improvement to the application experience metric for the online application that exceeds a threshold defined by a user-specified policy.
18. The apparatus as in claim 11 , wherein paths in the particular subset are associated with a common site.
19. The apparatus as in claim 11 , wherein the apparatus makes comparisons between the aggregated distributions for the different subsets of the set of paths based in part on predicted traffic loads for the different subsets.
20. A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute a process comprising:
predicting, by the device and for each of a set of paths via which traffic for an online application can be routed, a distribution of an application experience metric for the online application;
computing, by the device and for different subsets of the set of paths, aggregated distributions of their distributions of the application experience metric predicted by the device;
making, by the device, comparisons between the aggregated distributions for the different subsets of the set of paths; and
causing, by the device and based on the comparisons, the traffic for the online application to be routed via a particular subset of the set of paths.