Troubleshooting for Network Operators

6Root Cause Analysis and Definitive Remediation

“The spread of computers and the Internet”

Marc Andreessen

Chapter 6 presents a use-case for the root cause analysis and definitive remediation. Concretely, we present a root cause analysis mechanism using machine learning (ML) to identify the root causes of congestion, and then we implement an adaptive congestion control algorithm in the definitive module to solve it completely. These modules are described in detail in Figure 2.2.

6.1. Root cause analysis: machine learning based root cause analysis for SDN network

6.1.1. Introduction

Root cause analysis refers to a process of identifying and delimiting elements leading to anomalies. There are three main objectives in the root cause analysis (Mdini 2019):

– The first objective is to identify which network problems result in the anomalies (e.g. link failure, switch failure, etc.). If a network problem occurs, it will be classified to return the type of the problem (Fortes et al. 2015).

– The second objective is to identify events that lead to the anomalies. Event logs of network elements can be used to study their causalities and identify anomalous events (Nagaraj et al. 2012).

– The final objective is to localize network elements that result in the anomalies. Performance evaluation at network elements can be used to identify the anomalous elements (Dimopoulos et al. 2015).

The root cause analysis can return the type of network problems, anomalous events and anomalous network ...