Return and replacement protocol (RRP)
Systems and methods provide for managing faulting network devices. A first network device can receive an error. The first network device can generate one or more frames including data indicative of the error. The first network device can broadcast the one or more frames to one or more neighboring network devices. It may be determined that the first network device is inaccessible. The first data can be retrieved and presented from a second network device among the one or more neighboring network devices. In some embodiments, a network management system can utilize the first data to generate a machine learning model that classifies whether network devices are instances of network devices designated for a Return Merchandise Authorization (RMA) process. In some embodiments, the network management system can apply the first data to a machine learning classifier to determine whether to initiate the RMA process for the first network device.
1. A computer-implemented method, comprising:
receiving, by a first network device, a first error associated with the first network device;
generating, by the first network device, one or more first frames including first data indicative of the first error;
broadcasting, by the first network device, the one or more first frames to one or more neighboring network devices;
determining, after the broadcasting, that the first network device is unresponsive;
presenting the first data in response to a request to a second network device among the one or more neighboring network devices;
extracting one or more feature values from the one or more first frames, the one or more feature values including at least one of silicon aging data or silicon failure data of the first network device; and
determining a feature vector representation based on the one or more feature values.
2. The computer-implemented method of claim 1 , further comprising:
generating at least one of a naive Bayes, logistic regression, nearest neighbors, support vector machine, boosting, or neural network classifier based at least in part on the feature vector representation.
3. The computer-implemented method of claim 1 , further comprising:
applying the feature vector representation to one or more machine learning classifiers;
and
in response to the one or more machine learning classifiers indicating that the first network device is an instance of a class of network devices designated for Return Merchandise Authorization (RMA), initiating an RMA process for the first network device.
4. The computer-implemented method of claim 1 , further comprising:
applying the feature vector representation to one or more machine learning models; and
providing a prediction of a time that the first network device is expected to be designated for RMA to an instance of a class of network devices designated for RMA.
5. The computer-implemented method of claim 1 , further comprising:
receiving, by a network management system, the silicon aging data of the first network device; and
determining, by the network management system, a silicon failure metric of the first network device based at least in part on the silicon aging data.
6. The computer-implemented method of claim 1 , further comprising:
configuring, by a network management system, an interval at which a plurality of network devices transmit frames that include device error data to neighboring network devices.
7. The computer-implemented method of claim 1 , wherein the first frame is broadcast prior to loading of an operating system of the first network device.
8. The computer-implemented method of claim 1 , wherein the first frame is broadcast by a process given a highest process priority in the first network device.
9. The computer-implemented method of claim 1 , further comprising:
receiving, by the first network device from the second network device, one or more second frames including second data indicative of a second error associated with the second network device; and
presenting the second data in response to a request to the first network device.
10. A system, comprising:
one or more processors; and
memory including instructions that, when executed by the one or more processors, cause the system to:
receive, by a first network device of the system, a first error associated with the first network device;
generate, by the first network device, one or more first frames including first data indicative of the first error;
broadcast, by the first network device, the one or more first frames to one or more neighboring network devices;
determine, after the broadcast, that the first network device is unresponsive; and
present the first data in response to a request to a second network device among the one or more neighboring network devices;
extract one or more feature values from the one or more first frames, the one or more feature values including at least one of silicon aging data or silicon failure data of the first network device; and
determine a feature vector representation based on the one or more feature values.
11. The system of claim 10 , wherein the instructions, when executed, further cause the system to:
generate at least one of a naive Bayes, logistic regression, nearest neighbors, support vector machine, boosting, or neural network classifier based at least in part on the feature vector representation.
12. The system of claim 10 , wherein the instructions, when executed, further cause the system to:
apply the feature vector representation to one or more machine learning classifiers; and
in response to the one or more machine learning classifiers indicating that the first network device is an instance of a class of network devices designated for Return Merchandise Authorization (RMA), initiate an RMA process for the first network device.
13. The system of claim 10 , wherein the instructions, when executed, further cause the system to:
apply the feature vector representation to one or more machine learning models; and
provide a prediction of a time that the first network device is expected to be designated for RMA to an instance of a class of network devices designated for RMA.
14. A non-transitory computer-readable storage medium including instructions that, upon being executed by one or more processors of a system, cause the system to perform operations comprising:
receive, by a first network device of the system, a first error associated with the first network device;
generate, by the first network device, one or more first frames including first data indicative of the first error;
broadcast, by the first network device, the one or more first frames to one or more neighboring network devices;
determine, after the broadcast, that the first network device is unresponsive;
present the first data in response to a request to a second network device among the one or more neighboring network devices;
extract one or more feature values from the first frame, the one or more feature values including at least one of silicon aging data or silicon failure data of the first network device; and
determine a feature vector representation based on the one or more feature values.
15. The non-transitory computer-readable storage medium of claim 14 , wherein the instructions, when executed, further cause the system to:
receive the silicon aging data of the first network device; and
determine a silicon failure metric of the first network device based at least in part on the silicon aging data.
16. The non-transitory computer-readable storage medium of claim 14 , wherein the instructions, when executed, further cause the system to:
configure an interval at which a plurality of network devices of the system transmit frames that include device error data to neighboring network devices of the system.
17. The non-transitory computer-readable storage medium of claim 14 , wherein the first frame is broadcast prior to loading of an operating system of the first network device or the first frame is broadcast by a process given a highest process priority in the first network device.