Replacing static routing metrics with probabilistic models

1. A method comprising:

obtaining, by a device, a predictive model that predicts a behavior of a path in a network;

computing, by the device and based in part on the predictive model, a route in the network that includes the path, in accordance with a routing policy that instructs the device to use the predictive model as an attribute of the path during computation of the route and to use the predictive model only when computing routes for a particular type of traffic;

computing, by the device, another route in the network for a type of traffic other than the particular type of traffic, in accordance with a second routing policy that instructs the device to use static path metrics for the path instead of the predictive model during this computation;

validating, by the device, that the path exhibited the behavior predicted by the predictive model; and

initiating, by the device, retraining of the predictive model, when the behavior predicted by the predictive model does not match the behavior of the path.

2. The method as in claim 1 , wherein the network comprises a software-defined wide area network.

3. The method as in claim 1 , wherein the behavior of the path predicted by the predictive model is indicative of whether the path will violate a service level agreement (SLA) for traffic to be routed via the path, and wherein validating that the path exhibited the behavior predicted by the predictive model comprises:

determining whether the path violated the SLA as predicted by the predictive model based on information indicative of an actual performance of the path.

4. The method as in claim 1 , wherein validating that the path exhibited the behavior predicted by the predictive model comprises:

sending probes along the route, to collect probing statistics.

5. The method as in claim 1 , wherein validating that the path exhibited the behavior predicted by the predictive model comprises:

assigning a trust score to the predictive model based on a comparison between the behavior predicted by the predictive model and the behavior of the path.

6. The method as in claim 1 , wherein the predictive model is sent via the network as an Open Shortest Path First (OSPF) link state advertisement (LSA) or an Intermediate System to Intermediate System (IS-IS) type-length-value (TLV).

7. The method as in claim 1 , wherein computing the route in the network that includes the path comprises:

overriding a routing decision to use a different routing path, to test validity of the predictive model.

8. The method as in claim 1 , wherein the device comprises a router at an edge of the network.

9. 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:

obtain a predictive model that predicts a behavior of a path in a network;

compute, based in part on the predictive model, a route in the network that includes the path, in accordance with a routing policy that instructs the apparatus to use the predictive model as an attribute of the path during computation of the route and to use the predictive model only when computing routes for a particular type of traffic;

compute another route in the network for a type of traffic other than the particular type of traffic, in accordance with a second routing policy that instructs the device to use static path metrics for the path instead of the predictive model during this computation;

validate that the path exhibited the behavior predicted by the predictive model; and

initiate retraining of the predictive model, when the behavior predicted by the predictive model does not match the behavior of the path.

10. The apparatus as in claim 9 , wherein the network comprises a software-defined wide area network.

11. The apparatus as in claim 9 , wherein the behavior of the path predicted by the predictive model is indicative of whether the path will violate a service level agreement (SLA) for traffic to be routed via the path, and wherein the apparatus validates that the path exhibited the behavior predicted by the predictive model by:

determining whether the path violated the SLA as predicted by the predictive model based on information indicative of an actual performance of the path.

12. The apparatus as in claim 9 , wherein the apparatus validates that the path exhibited the behavior predicted by the predictive model by:

sending probes along the route, to collect probing statistics.

13. The apparatus as in claim 9 , wherein the apparatus validates that the path exhibited the behavior predicted by the predictive model by:

assigning a trust score to the predictive model based on a comparison between the behavior predicted by the predictive model and the behavior of the path.

14. The apparatus as in claim 9 , wherein the predictive model is sent via the network as an Open Shortest Path First (OSPF) link state advertisement (LSA) or an Intermediate System to Intermediate System (IS-IS) type-length-value (TLV).

15. The apparatus as in claim 9 , wherein the apparatus computes the route in the network that includes the path by:

overriding a routing decision to use a different routing path, to test validity of the predictive model.

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

obtaining, by the device, a predictive model that predicts a behavior of a path in a network;

computing, by the device and based in part on the predictive model, a route in the network that includes the path, in accordance with a routing policy that instructs the device to use the predictive model as an attribute of the path during computation of the route and to use the predictive model only when computing routes for a particular type of traffic;

computing, by the device, another route in the network for a type of traffic other than the particular type of traffic, in accordance with a second routing policy that instructs the device to use static path metrics for the path instead of the predictive model during this computation;

validating, by the device, that the path exhibited the behavior predicted by the predictive model; and

initiating, by the device, retraining of the predictive model, when the behavior predicted by the predictive model does not match the behavior of the path.