Ashish Anand is the Director of Global Networks at Marriott International.
Network infrastructures that can be relied upon 24/7 are the backbone of any modern digital enterprise, business continuity, user experience or cybersecurity strategy. However, network management has become more complex as IT architectures spread across hybrid clouds, edge computing and Internet of Things (IoT) endpoints and with demands to maintain real-time responsiveness. Downtime not only affects productivity, but it also leads to substantial financial and reputational damage.
As enterprise networks have matured, their complexity has increased to the extent that human intervention, troubleshooting and recovery are no longer viable. Monitoring solutions may generate alerts, but network experts are required to diagnose the problem and drive the resolution. Even with ML/AI-driven analyses and solutions, network remediation is still driven by the operator, resulting in relatively high time to recovery (MTTR).
Network intent has already emerged as a common basis for managing networks automatically, but there remains a significant gap between this high-level intent and the low-level corrective actions taken.
Self-healing network infrastructures are next-generation autonomous networks that use AI-based intent recognition to not only detect failures but also predict disruptions before they impact the business and pre-emptively remediate the issue without any human intervention.
Let’s take a closer look at how self-healing frameworks use AI-based intent recognition systems.
What Is AI-Based Intent Recognition?
Intent-based networking (IBN) is a type of software-defined networking (SDN) that allows users to define and manage a network by describing what the user wants the network to do (also known as the intent) instead of configuring every layer and device.
Intent on networking refers to the user’s goals that are translated into action to ensure they are reliably enforced so that network services can function at expected levels. AI-based intent recognition uses NLP, ML and policy-based analytics to understand business objectives, translate objectives into actionable network configurations and continuously monitor performance to ensure compliance with the intent (feedback loop).
This is very powerful as even users without much in-depth network knowledge can express their intent for behavior.
How Self-Healing Networks Work
A self-healing network is one that behaves like an immune system in the human body and can detect anomalies autonomously, diagnose the root cause and initiate corrective actions to resolve issues without any human intervention. The goal is to keep the network running smoothly with minimal downtime, even in the face of hardware failures, configuration errors or other issues.
The structure of self-healing networks features five components:
• Monitoring: Continuously ingest and collect logs, metrics and traces from all devices.
• Anomaly Detection: Use AI models to detect deviations from normal behavior (e.g., unusual latency spikes, packet losses or topology changes).
• Intent Correlation: Correlate the anomaly against the system objectives, high-level policy goals (intent) to check whether the network is still meeting required intent (e.g., QoS for video conferencing).
• Autonomous Remediation: Automatically triggers actions such as rerouting traffic and applying security patches to autonomously remediate the issue.
• Learning Loop: Feedback is used to refine future predictions and responses.
This could be implemented by automating common network recovery tasks. It’s important to redesign your network with redundant paths (via routing protocols like BGP, OSPF) and failover systems. You should also adopt and deploy SDN platforms with centralized control to help with your automation efforts and self-healing.
Next, you’ll need to integrate platforms with AI-based insights. In large enterprises, explore intent-based networking solutions that align network behavior with business policies. Begin with one self-healing use case (e.g., restarting a down VPN, rerouting traffic when a link fails, etc.), then expand it to more critical use cases over time.
System Architecture
The framework’s architecture is made up of four main layers:
• Intent Learning Layer: This layer trains AI/ML models on DevOps CI/CD pipelines and IaC configuration templates (e.g., Terraform, Ansible scripts). It also extracts and defines “intent” by mapping high-level policies (e.g., redundancy, latency SLAs, firewall rules) into machine-readable network goals.
• Telemetry Monitoring Layer: This layer normalizes all heterogeneous telemetry into a unified, multimodal data model.
• Drift Detection And Root Cause Prediction Layer: This layer compares live telemetry with the “intent baseline.” It uses anomaly detection and causal inference models to identify drift. It then predicts the likely cause from a range of potential misconfigurations, capacity bottlenecks or policy conflicts.
• Healing Orchestration Layer: This layer automatically generates and provisions corrective IaC scripts aligned with original DevOps workflows (e.g., rolling back configs, adjusting routing policies). It performs closed-loop remediation validation by re-checking the state after remediation.
Real-World Applications
Some real-world applications include the telecom sector (predictive rerouting can ensure mobile connectivity is always on), financial institutions (ensuring low-latency and secure transactions), healthcare (enabling uptime for telemedicine and patient monitoring systems) and hospitality (enhanced availability by auto-scaling based on increased demand.
The real-world applications extend beyond these use cases and will help redefine how networking is architected and operated across enterprises.
Challenges And Considerations
While AI-driven self-healing will be able to provide the automated detection and resolution of faults without human intervention, thus driving down OPEX costs, reducing downtime with less service disruption and improving operational efficiency as a result of less time spent on troubleshooting and eliminating repetitive tasks, there are some challenges to consider.
Enterprises need explainable AI and transparency in algorithms to validate remediation actions. Sensitive telemetry data must be protected and properly secured. It also needs to be able to easily integrate legacy systems and equipment with intent-based frameworks. Lastly, there must be an ability to allow administrators to override self-healing actions.
The Road Ahead
AI, intent-based networking and automation will set the foundation for next-generation fully autonomous networks in which intelligence is built in from the start (i.e., where compute resources and underlying infrastructure no longer need to be manually configured and managed).
Network architecture is expected to change as enterprises move to self-healing infrastructures. They will shift from a focus on reactive management to more predictive and proactive orchestration to better ensure resilience, security and performance in an increasingly digital world.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
link