Predictive network routing with dynamic smoothing envelope creation for noisy network timeseries

1. A method comprising:

generating, by a device, a plurality of smoothed timeseries by applying smoothing envelopes of different durations to a timeseries of a path metric for a path in a network that is used to convey traffic of an online application;

using, by the device, the plurality of smoothed timeseries and the timeseries of the path metric to make predictions as to whether the path will provide an unacceptable user experience in the online application;

selecting, by the device, a smoothing envelope of a particular duration, by comparing performance metrics for the predictions; and

using, by the device, a timeseries of the path metric smoothed using the smoothing envelope of the particular duration to make predictive routing decisions in the network for the traffic of the online application.

2. The method as in claim 1 , wherein the performance metrics for the predictions comprise precision metrics or recall metrics for the predictions.

3. The method as in claim 1 , wherein the path metric indicates a probability of the path violating a service level agreement associated with the online application.

4. The method as in claim 1 , wherein the different durations are selected based in part on a processing time associated with making the predictions.

5. The method as in claim 1 , further comprising:

providing, by the device, an indication of the smoothing envelope of the particular duration to a user interface; and

receiving, at the device and from the user interface, an instruction to make the predictive routing decisions using the timeseries of the path metric smoothed using the smoothing envelope of the particular duration instead of the timeseries of the path metric.

6. The method as in claim 1 , wherein generating the plurality of smoothed timeseries comprises:

identifying the particular duration by performing timelagged cross correlation between the timeseries of the path metric and a timeseries of a second path metric.

7. The method as in claim 1 , wherein generating the plurality of smoothed timeseries comprises:

determining a distribution difference between the path metric and a second path metric.

8. The method as in claim 1 , further comprising:

repeatedly evaluating whether a smoothing envelope having a different duration than that of the particular duration should be used to make the predictive routing decisions.

9. The method as in claim 1 , further comprising:

using a prediction model to output the smoothing envelope of the particular duration.

10. The method as in claim 1 , wherein the predictive routing decisions comprise a decision to reroute the traffic of the online application in the network, in advance of a prediction that the path will provide unacceptable user experience in the online 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:

generate a plurality of smoothed timeseries by applying smoothing envelopes of different durations to a timeseries of a path metric for a path in a network that is used to convey traffic of an online application;

use the plurality of smoothed timeseries and the timeseries of the path metric to make predictions as to whether the path will provide an unacceptable user experience in the online application;

select a smoothing envelope of a particular duration, by comparing performance metrics for the predictions; and

use a timeseries of the path metric smoothed using the smoothing envelope of the particular duration to make predictive routing decisions in the network for the traffic of the online application.

12. The apparatus as in claim 11 , wherein the performance metrics for the predictions comprise precision metrics or recall metrics for the predictions.

13. The apparatus as in claim 11 , wherein the path metric indicates a probability of the path violating a service level agreement associated with the online application.

14. The apparatus as in claim 11 , wherein the different durations are selected based in part on a processing time associated with making the predictions.

15. The apparatus as in claim 11 , wherein the process when executed is further configured to:

provide an indication of the smoothing envelope of the particular duration to a user interface; and

receive, from the user interface, an instruction to make the predictive routing decisions using the timeseries of the path metric smoothed using the smoothing envelope of the particular duration instead of the timeseries of the path metric.

16. The apparatus as in claim 11 , wherein the apparatus generates the plurality of smoothed timeseries by:

identifying the particular duration by performing timelagged cross correlation between the timeseries of the path metric and a timeseries of a second path metric.

17. The apparatus as in claim 11 , wherein the apparatus generates the plurality of smoothed timeseries by:

determining a distribution difference between the path metric and a second path metric.

18. The apparatus as in claim 11 , wherein the process when executed is further configured to:

repeatedly evaluate whether a smoothing envelope having a different duration than that of the particular duration should be used to make the predictive routing decisions.

19. The apparatus as in claim 11 , wherein the predictive routing decisions comprise a decision to reroute the traffic of the online application in the network, in advance of a prediction that the path will provide unacceptable user experience in the online application.

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

generating, by the device, a plurality of smoothed timeseries by applying smoothing envelopes of different durations to a timeseries of a path metric for a path in a network that is used to convey traffic of an online application;

using, by the device, the plurality of smoothed timeseries and the timeseries of the path metric to make predictions as to whether the path will provide an unacceptable user experience in the online application;

selecting, by the device, a smoothing envelope of a particular duration, by comparing performance metrics for the predictions; and

using, by the device, a timeseries of the path metric smoothed using the smoothing envelope of the particular duration to make predictive routing decisions in the network for the traffic of the online application.