Reactive assurance models are being pushed to the limit as AI-driven traffic rips through the network
Networks are increasingly carrying cutting-edge AI-related workloads, and that is straining traditional network testing approaches in unexpected ways.
On one hand, new AI functions and applications are being loaded into the network to operationalize AI, and on the other hand, new AI-related use cases are mushrooming at the user end, demanding faster, more reliable connectivity.
A common misconception is that the existing ways of testing and measuring network health and performance could be carried over to these AI-enriched or AI-native networks. But increasingly the older model is cracking under the weight of AI’s enormous scale and specific traffic profiles.
“Legacy assurance is often too disconnected from the speed and complexity that we’re seeing in modern networks,” said Ross Cassan, senior director of assurance strategy at Spirent during a recent RCR webinar on AI-era network test, measurement, and assurance.
Cassan argues that assurance frameworks built on top of older models are doomed to fail in the AI-era. Here’s why:
Fragmented data
Data in telecommunications is distributed across legacy systems, cloud platforms, and disparate departments. This fragmentation obstructs a 360-degree view of user experience – a must to assure the network – not to mention, takes a heavy toll on operational efficiency. This could lead to a series of unfavorable outcomes, including service delays, security risks, increased OpEx, and ultimately poor customer experience resulting from the blind spots.
The fragmentation is not only confined to data, but runs across tools, workflows, teams, and departments, leaving current assurance processes domain-specific, periodic, and stuck in silos like the ecosystems they are designed to assure. It is no surprise that they are proving inadequate to prevent service-level agreement (SLA) breaches in dynamic network environments.
Network Intelligence
Organizations need data delivered to them as insights to act on them. That’s 101 of assurance. While it is the primary goal of all service assurance frameworks to bring these insights to the fingertips of engineers as quickly as possible, current root cause analysis (RCA) processes are slow because of the aforementioned fragmentation, leading to longer outages, revenue leakage, and customer churn.
In order to obtain real-time intelligence, operators require smart AI and machine learning algorithms that can crunch through volumes of data and correlate them in minutes, delivering that holistic visibility that allows for deeper understanding of what needs to be upgraded, how to ensure better quality of experience (QoE), how to prevent service degradation, etc. in a predictive manner.
Once again, this requires agentic AI workflows to be embedded in assurance frameworks to enable swift root cause analysis and instant insights.
Reactive processes
The biggest reason why service assurance needs a step change in the AI era is because today the discipline is largely reactive, meaning it finds issues and failures, implements corrective actions, and remedies the situation after the fact. “It tells you what happened after a problem occurred, but not always fast enough to prevent a customer impact,” added Cassan.
AI-driven applications are less forgiving to such performance disruptions. The smallest of hiccups can lead to annoyingly slow response times, inconsistent outputs, and even application failures in mission-critical scenarios.
Operators need to identify risks and put controls in place before they turn into failures, so that issues can be resolved before they reach the users and hurt their experience.
To learn more, watch the webinar for free on RCRwireless.com.
