Table of Contents
As networks grow heavier and more complex with addition of AI applications, testing and validation approaches are evolving to keep up
Artificial intelligence is increasingly embedded in modern network infrastructures. A roll of new use cases have emerged in the recent years, making room for AI to be implemented at every layer and detail of the network to augment it from within.
Broadly, implementation is happening at two levels: For automating network operations, management, and security. This is where network equipment manufacturers like Cisco, HPE Aruba, and Juniper Networks are steadily incorporating AI capabilities and controls into network gears — switches, routers, access points, firewalls, gateways, radio equipment — to predict and prevent attacks, automate policy and configuration management, deliver intent-based networking with low human touch, and so on.
And at the infrastructure level, operators and service providers are leveraging AI to upgrade and optimize the network infrastructure across wired, wireless and data center networks. AI workloads demand extreme throughout, ultra low-latency, and lossless connectivity from edge to core. To guarantee that, providers are integrating AI in all parts of the network. In radio access networks (RAN), AI is helping maximize infrastructure utilization, spectral efficiency, and throughput and handover speeds, while unlocking new levels of energy efficiency. In networking, AI assistants are working side by side with human administrators, expediting functions like policy control, pattern analysis, and troubleshooting, making network management simpler and low-touch. And AI-based predictive traffic optimization and load balancing are ensuring optimum user experience by minimizing lag and downtime.
“We are seeing a raft of AI coming into the networks,” said Stephen Douglas, head of marketing at Spirent, echoing the trend in a webinar with RCR Wireless News.
According to Gartner, AI spending is soon to hit $2 trillion worldwide. A bulk of that will be driven by investments in the underlying infrastructure. This checks out with what companies are witnessing on the ground.
“A number of the service providers that we work with at the moment are exploring the types of optimizations and upgrades in the network infrastructure that they need to support AI traffic, whether that’s new types of capacities and throughputs through to how they can support segmentation on the wireline, the fixed, and the wireless network,” Douglas told.
But while AI is key to optimizing network performance, an AI-enhanced network needs active and continuous evaluation to ensure that it’s performing better than a legacy network.
“We are seeing a demand across the board for continuous validation in the lab and the live environment of those AI capabilities in the vendor equipment. Are they fit for purpose? Are they delivering the relevant types of responses?” said Douglas.
Additionally, testing is required to ensure that the AI is not bringing in hidden risks that could make the network vulnerable to attacks.
Testing approaches for an AI-enhanced network
There are some testing capabilities that particularly come handy for evaluating complex AI-enriched networks.
Network Digital Twin — A digital twin is a virtual replica of the operational network. This version is created as a test bed for performing continuous testing and assurance on the network. What’s interesting with this twin network is that operators can test on demand without interrupting the living network. Any aspect of network operations, whether that’s traffic congestion, equipment failure, downtime, change management, optimization, or security breaches, can be examined on this replica network.
Synthetic Test — A simulation-based testing approach, synthetic monitoring entails generating network traffic data for any and all scenarios and injecting the traffic into the live network. This allows for stress-testing or evaluating the readiness of the network. The technique is often leveraged to verify quality of service (QoS), prepare for demand spikes, and ensure an even experience for all users.
Continuous and Active Testing — Another proactive approach, continuous and active monitoring sits at the core of network test and assurance. It involves holistic testing of the network end to end, not only in lab environments but also in live environments. Continuous probing uncovers real-time performance analytics that allow operators to see and resolve service degradations ahead of time, thus averting end-user experience issues.
As service providers tap AI to optimize network performance and user satisfaction, test and assurance unlocks opportunities for efficiency, and improved service delivery.
