Root Cause Analysis of Optical Transport Network Channel Failure
No Thumbnail Available
Date
2024-08
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
As demands for high bandwidth surge due to 4K/8K streaming, 5G networks, and
cloud-based applications, robust optical transport networks (OTNs) are critical. OTN failures
can significantly impact service quality, network availability, and service level agreements
(SLAs). While research has addressed fault prediction and localization, a gap exists
in root cause analysis (RCA) for OTN channel failures. This study proposes a novel
approach utilizing machine learning (ML) to pinpoint the root cause of these failures efficiently.
This study compared four ML classifier models to analyze real data from ethio telecom’s
network. The data included eight key features that influence OTN channel performance.
Extreme Gradient Boosting (XGBoost) emerged as the superior performer,
achieving an impressive 99.91% accuracy and a high F1-score of 97.5%. Furthermore, it
excelled in efficiency, with training times of just 5.42 seconds and testing times of 0.2 seconds.
Interestingly, the model identified minimum input optical power (Min IOP) as the
most critical factor, suggesting that extrinsic loss within the fiber optic cables is a major
cause of OTN channel failures.
This study explores a novel ML system for OTN RCA, enabling faster and more
precise root cause identification. This empowers network operators to proactively address
issues and ensure optimal performance, significantly boosting network reliability and efficiency
in the high-bandwidth age.
Description
Keywords
OTN, RCA, Channel Failure, Feature Importance, XGBoost, Min IOP, SM Errored Second, Extrinsic loss