Link fault management for optical adapters
An apparatus for optical link fault management includes a processor in a network device with an optical adapter of an optical link, and a memory that stores program code. The program code is executable by the processor to store state parameters from the optical adapter in a data log, input the state parameters from the data log into a failure prediction model, and generate, using the failure prediction model, a probability of failure of the optical adapter based on current state parameters from the data log. In response to the probability of failure reaching a failure threshold, the program code is executable by the processor to decrease data traffic in the optical link and to send an alert comprising the probability of failure. The failure threshold includes a probability of failure indicative of impending failure of the optical adapter.
1. An apparatus comprising:
a processor in a network device comprising an optical adapter of an optical link; and
a memory that stores program code executable by the processor to:
store state parameters from the optical adapter in a data log;
input the state parameters from the data log into a failure prediction model;
generate, using the failure prediction model, a probability of failure of the optical adapter based on current state parameters from the data log;
monitor data traffic patterns for the optical link;
estimate, based on the data traffic patterns, a time for corrective action for the optical adapter of the optical link, the time for corrective action comprising a time prior to a predicted failure based on the probability of failure, the estimated time for the corrective action comprising a time with a low amount of disruptiveness to data traffic; and
in response to the probability of failure reaching a failure threshold:
decrease data traffic in the optical link; and
send an alert comprising the probability of failure,
wherein the failure threshold comprises a probability of failure indicative of impending failure of the optical adapter, and wherein the alert comprises the time for the corrective action.
2. The apparatus of claim 1 , wherein the program code to monitor data traffic patterns for the optical link comprises determining a pattern comprising a time period of when data traffic to the optical link is low.
3. The apparatus of claim 1 , wherein decreasing data traffic in the optical link comprises increasing a cost of using the optical link, wherein increasing the cost of using the optical link decreases data traffic to the optical link and increases data traffic to a redundant optical link.
4. The apparatus of claim 1 , wherein the program code is further executable by the processor to:
gather failure data of a plurality of optical adapters and related state parameters of the plurality of optical adapters; and
refine, using machine learning, the failure prediction model based on the failure data of the plurality of optical adapters and related state parameters of the plurality of optical adapters.
5. The apparatus of claim 4 , wherein the network device comprises the failure prediction model and a global network manager comprises an additional failure prediction model, wherein the gathered failure data of the plurality of optical adapters and related state parameters of the plurality of optical adapters is gathered at the global network manager which refines the failure prediction model at the global network manager, and wherein the program code is further executable by the processor to receive model parameters from the global network manager and to modify the failure prediction model at the network device based on the refinements to the failure prediction model at the global network manager.
6. The apparatus of claim 1 , wherein the program code is further executable by the processor to determine a current link health, wherein the current link health is a function of the probability of failure of the optical adapter and a frame checksum error rate.
7. The apparatus of claim 1 , wherein the program code is further executable by the processor to determine a current aggregated link health (“LHA”):
LHA
=
max
(
0
,
100
-
(
E
r
L
t
h
·
100
)
)
wherein:
E r is a frame checksum error rate, and
L th is a frame checksum error threshold.
8. The apparatus of claim 7 , wherein the program code is further executable by the processor to determine a predicted aggregated link failure time comprising a predicted time when the LHA reaches zero.
9. The apparatus of claim 7 , wherein the program code is further executable by the processor to determine a link alert time comprising a predicted time when the LHA reaches an LHA threshold.
10. The apparatus of claim 1 , wherein the program code is further executable by the processor to determine a predicted link health alert time T LHF comprising a predicted time of the probability of failure of the optical adapter reaching a probability of failure threshold independent of a frame checksum error rate.
11. The apparatus of claim 1 , wherein the program code is further executable by the processor to determine a predicted link efficiency failure time T LEF (L TH ) comprising a time when a frame checksum error rate of the optical adapter is predicted to reach a frame checksum error threshold independent of the probability of failure.
12. A method comprising:
monitoring, by use of a processor, data traffic patterns for an optical link comprising an optical adapter;
storing, by use of the processor, state parameters from the optical adapter in a data log;
inputting, by use of the processor, the state parameters from the data log into a failure prediction model;
generating, by use of the processor and using the failure prediction model, a probability of failure of the optical adapter based on current state parameters from the data log;
estimating, by use of the processor and based on the data traffic patterns, a time for corrective action for the optical adapter, the time for corrective action comprising a time prior to a predicted failure based on the probability of failure; and
sending, by use of the processor, an alert comprising the probability of failure in response to the probability of failure reaching a failure threshold, the failure threshold comprising a probability of failure indicative of impending failure of the optical adapter, wherein the alert comprises an estimated time for the corrective action, the estimated time for the corrective action comprises a time with a low amount of disruptiveness to data traffic.
13. The method of claim 12 , further comprising, in response to the probability of failure reaching the failure threshold, decreasing data traffic in the optical link.
14. The method of claim 13 , wherein decreasing data traffic in the optical link comprises increasing a cost of using the optical link, wherein increasing the cost of using the optical link decreases data traffic to the optical link and increases data traffic to a redundant optical link.
15. The method of claim 12 , further comprising:
gathering failure data of a plurality of optical adapters and related state parameters of the plurality of optical adapters; and
refining, using machine learning, the failure prediction model based on the failure data of the plurality of optical adapters and related state parameters of the plurality of optical adapters,
wherein the failure prediction model is in a network device and a global network manager comprises an additional failure prediction model, wherein the gathered failure data of the plurality of optical adapters and related state parameters of the plurality of optical adapters are gathered at the global network manager which refines the failure prediction model at the global network manager, and further comprising receiving model parameters from the global network manager and modifying the failure prediction model at the network device based on the refinements to the failure prediction model at the global network manager.
16. The method of claim 12 , wherein monitoring data traffic patterns for the optical link comprises determining a pattern comprising a time period of when data traffic to the optical link is low.
17. The method of claim 12 , further comprising determining a current aggregated link health (“LHA”):
LHA
=
max
(
0
,
100
-
(
E
r
L
t
h
·
100
)
)
wherein:
E r is a frame checksum error rate, and
L th is a frame checksum error threshold, and
further comprising determining a predicted aggregated link failure time comprising a predicted time when the LHA reaches zero.
18. A program product comprising a non-transitory computer readable storage medium with program code, the program code being configured to be executable by a processor to perform operations comprising:
storing state parameters from an optical adapter of an optical link in a data log;
inputting the state parameters from the data log into a failure prediction model;
generating, using the failure prediction model, a probability of failure of the optical adapter based on current state parameters from the data log;
monitoring data traffic patterns for the optical link;
estimating, based on the data traffic patterns, a time for corrective action for the optical adapter of the optical link, the time for corrective action comprising a time prior to a predicted failure based on the probability of failure, the estimated time for the corrective action comprising a time with a low amount of disruptiveness to data traffic; and
in response to the probability of failure reaching a failure threshold:
decreasing data traffic in the optical link; and
sending an alert comprising the probability of failure, the failure threshold comprising a probability of failure indicative of impending failure of the optical adapter, wherein the alert comprises the time for the corrective action.