Discovery of port-to-port connectivity across network layers using machine learning
Systems and methods for discovering the connectivity topology of cross-layer links in a multi-layer network are provided. In one implementation, a method includes a step of obtaining input data related to a plurality of Network Elements (NEs) operating in a plurality of layers within a multi-layer network having one or more cross-layer port-to-port connectivity links therebetween. The method also includes the step of utilizing Machine Learning (ML) processes and the input data to discover the one or more cross-layer port-to-port connectivity links between pairs of NEs operating in different layers of the plurality of layers within the network.
1. A Network Management System (NMS) comprising
a processing device, and
a memory device configured to store instructions that, when executed, cause the processing device to
obtain input data related to a plurality of Network Elements (NEs) operating in a plurality of layers within a network having a cross-layer topology of one or more cross-layer port-to-port connectivity links therebetween, and
utilize a Machine Learning (ML) model and the input data to discover the cross-layer topology including the one or more cross-layer port-to-port connectivity links between pairs of NEs operating in different layers of the plurality of layers within the network,
wherein the ML model is trained utilizing one or more ML processes with multi-variate time-series data including a known cross-layer topology with known cross-layer port-to-port connectivity links therein.
2. The NMS of claim 1 , wherein the instructions, when executed, further cause the processing device to format the input data by performing a data aggregation process, wherein the data aggregation process includes one or more of data synchronization and data concatenation.
3. The NMS of claim 1 , wherein the instructions, when executed, further cause the processing device to utilize the ML model to discover connectivity topology between multiple pairs of NEs operating in the different layers within the network using a data-pattern matching process.
4. The NMS of claim 1 , wherein the instructions, when executed, further cause the processing device to utilize the ML model to reconcile an inventory database by one of
population of a new inventory in the inventory database, and
verification of accuracy of an existing inventory in the inventory database and modification of the existing inventory based on the verification.
5. A non-transitory computer-readable medium configured to store computer logic having instructions that, when executed, cause one or more processing devices to
obtain input data related to a plurality of Network Elements (NEs) operating in a plurality of layers within a network having a cross-layer topology of one or more cross-layer port-to-port connectivity links therebetween, and
utilize a Machine Learning (ML) model and the input data to discover the cross-layer topology including the one or more cross-layer port-to-port connectivity links between pairs of NEs operating in different layers of the plurality of layers within the network,
wherein the ML model is trained utilizing one or more ML processes with multi-variate time-series data including a known cross-layer topology with known cross-layer port-to-port connectivity links therein.
6. The non-transitory computer-readable medium of claim 5 , wherein the instructions, when executed, cause the one or more processing devices to
train the ML model using inventory data from a controlled environment where port-to-port connectivity information of the inventory data is known to be correct, and
label training and testing data from which the ML model is derived,
wherein the controlled environment includes one of a lab, a reference production network, and a subset of the network.
7. The non-transitory computer-readable medium of claim 5 , wherein the instructions, when executed, cause the one or more processing devices to estimate precision and recall information based on results from utilizing the ML model.
8. The non-transitory computer-readable medium of claim 5 , wherein the different layers include an optical layer overlaid by an Internet Protocol (IP) layer and the input data includes multi-variate time-series Performance Metrics (PM) of the plurality of NEs.
9. A method comprising the steps of
obtaining input data related to a plurality of Network Elements (NEs) operating in a plurality of layers within a network having a cross-layer topology of or more cross-layer port-to-port connectivity links therebetween, and
utilizing a Machine Learning (ML) model and the input data to discover the cross-layer topology including the one or more cross-layer port-to-port connectivity links between pairs of NEs operating in different layers of the plurality of layers within the network,
wherein the ML model is trained utilizing one or more ML processes with multi-variate time-series data including a known cross-layer topology with known cross-layer port-to-port connectivity links therein.
10. The method of claim 9 , further comprising the step of utilizing the ML model to discover connectivity topology between multiple pairs of NEs operating in the different layers within the network.
11. The method of claim 9 , further comprising the step of formatting the input data by performing a data aggregation process.
12. The method of claim 11 , wherein the data aggregation process includes one or more of data synchronization and data concatenation.
13. The method of claim 9 , wherein discovering the one or more cross-layer port-to-port connectivity links includes a data-pattern matching process.
14. The method of claim 9 , further comprising the step of reconciling an inventory database.
15. The method of claim 14 , wherein the step of reconciling the inventory database includes one of
populating a new inventory in the inventory database, and
verifying accuracy of an existing inventory in the inventory database and modifying the existing inventory based on the verifying.
16. The method of claim 9 , wherein the one or more ML processes include:
training the ML model using inventory data from a controlled environment where port-to-port connectivity information of the inventory data is known to be correct, and
labelling training and testing data from which the ML model is derived,
wherein the controlled environment includes one of a lab, a reference production network, and a subset of the network.
17. The method of claim 9 , further comprising the step of estimating precision and recall information based on results from utilizing the ML model.
18. The method of claim 9 , wherein the different layers include an optical layer overlaid by an Internet Protocol (IP) layer, and wherein the method is performed independently of network protocols associated with the optical layer and IP layer.
19. The method of claim 9 , wherein the input data includes multi-variate time-series Performance Metrics (PMs) of the plurality of NEs.
20. The method of claim 9 , further comprising the steps of
modifying one or more ports of the one or more NEs during a maintenance window, wherein modifying the one or more ports includes one of:
inducing a specific traffic pattern,
altering physical characteristics of the one or more ports, and
turning off the one or more ports,
observing an effect on other NEs, and
tuning a pattern recognition process of the ML model to improve the precision of the pattern recognition process.