Pro-active Performance Monitoring in Optical Networks using Frequency Aware Seq2Seq Model
Rishabh Jain1, Umesh Sajjanar2
1Rishabh Jain†, Research Associate, Media and Data Science Research Lab, Adobe, Noida, India.
†Work done while working at Cisco AIR Lab
2Umesh Sajjanar, Engineering Manager, Cisco Artificial Intelligence Research Lab, Cisco Systems Inc, ORR Bangalore 560103 India.
Manuscript received on 07 December 2022 | Revised Manuscript received on 26 December 2022 | Manuscript Accepted on 15 February 2023 | Manuscript published on 28 February 2023 | PP: 1-10 | Volume-3 Issue-2 February 2023 | Retrieval Number: 100.1/ijdcn.B5028023223 | DOI: 10.54105/ijdcn.B5028.023223
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Published by Lattice Science Publication (LSP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Performance Monitoring (PM) and Fault Detection have always been a reactionary approach in Optical Networks for most service providers. Any kind of fault (power surge, ageing issues, equipment faults and failures, natural calamities, etc.) in an optical network is detected only after the fault has occurred and mitigation is performed afterward. The resultant service outages for end-users cause huge financial and reputation losses to the vendors. Therefore, there is a strong need for proactive detection of faults to limit disruption and provide uninterrupted services to clients. We achieve this objective by doing a multi-horizon time series prediction of Bit Error Rate at the receiver end of an optical circuit using our custom designed Frequency aware Sequence to Sequence (FaS2S) Neural Network. The predicted value of BER can be used to notify users of failure scenarios before they occur. Further corrective action, such as automatic re-routing or manual intervention can then be taken by the user. With this model, we can even configure the network properties dynamically during periods of low BER to push the network efficiency to its maximum capacity. See inference Video for BER inference capabilities of FaS2S Keywords: Performance Monitoring, Optical Networks, Artificial IntelligencePerformance Monitoring (PM) and Fault Detection have always been a reactionary approach in Optical Networks for most service providers. Any kind of fault (power surge, ageing issues, equipment faults and failures, natural calamities, etc.) in an optical network is detected only after the fault has occurred and mitigation is performed afterward. The resultant service outages for end-users cause huge financial and reputation losses to the vendors. Therefore, there is a strong need for proactive detection of faults to limit disruption and provide uninterrupted services to clients. We achieve this objective by doing a multi-horizon time series prediction of Bit Error Rate at the receiver end of an optical circuit using our custom designed Frequency aware Sequence to Sequence (FaS2S) Neural Network. The predicted value of BER can be used to notify users of failure scenarios before they occur. Further corrective action, such as automatic re-routing or manual intervention can then be taken by the user. With this model, we can even configure the network properties dynamically during periods of low BER to push the network efficiency to its maximum capacity. See inference Video for BER inference capabilities of FaS2S.
Keywords: Performance Monitoring, Optical Networks, Artificial Intelligence.
Scope of the Article: Optical and High-Speed Access Networks