English
AI is suddenly everywhere, but its impacts are only beginning to be felt. IDC estimates spending on AI software will reach $632B with a CAGR of 29% by 2028.
Communications networks are benefiting from AI as new capabilities enhance, automate, and optimize network management, performance, and security.
At the same time, AI workloads are pushing networks to the limit with relentless traffic, processing, and resiliency demands, requiring seamless uplink and downlink access to support the real-time data processing and inference needs of cutting-edge applications. These networks were not originally designed for AI’s need for data-intensive, high-bandwidth, low-latency, and lossless connectivity between distributed edge nodes and cloud data centers.
The impact is so dramatic that it is already driving a complete re-architecture of data centers. Dell’Oro projects data center Capex will increase from $260 billion in 2023 to more than $500 billion in 2028.
Wireless and wireline service providers are also impacted, rushing to upgrade network infrastructure to meet the needs of AI applications while simultaneously enhancing networks with AI intelligence. On the latter, they are proceeding cautiously as they build confidence and trust levels around AI-based decision making, sensitive data handling, and security.
Adopting new testing strategies and approaches is essential to address emerging AI challenges and validate that networks built for and enhanced with AI perform as expected. New methodologies like realistic AI workload emulation and network digital twins are speeding comprehensive and reproducible testing of AI at scale and under congestion, without requiring investment in expensive, proprietary labs.
These approaches are the subject of Spirent’s latest eBook, Testing Networks Built for and Enhanced with AI, which explores how AI’s requirements are impacting current networks and the latest trends around AI’s incorporation into critical network functions. It also examines how test and assurance addresses AI adoption challenges and the invaluable role they play in validating AI use cases. Insights from Spirent’s customer engagements and market experience provide a window into how new testing approaches are already benefiting early adopters.
How AI impacts networks
Today’s networks, designed primarily for video and internet traffic, are being optimized for AI communications, processes, and performance requirements, impacting the entire network.
AI is being tasked with addressing increased traffic generated by AI applications while also supporting new workloads, fundamentally reshaping network demands. Whether it’s distributing massive amounts of training data across compute nodes and GPUs, running inference models to deliver split-second predictions, or managing the intense traffic generated by AI-driven applications, there is an increasing need for high-performance, low-latency connectivity.
Training and inference rely on seamless communication between nodes to avoid delays, while AI applications introduce new traffic patterns that push networks to their limits. To meet these challenges, networks must evolve to ensure millisecond-level responsiveness, accuracy, and the capacity to handle rapidly growing data volumes.
Data centers are investing in new transport and interconnect architectures to deliver the bandwidth and performance necessary for AI traffic, with new networking capabilities and protocols being implemented to optimize these processes. Advanced Ethernet protocols are being adopted, as a migration to 800G and eventually 1.6 Terabit interconnect speeds begin to ensure networks can accommodate AI’s data-intensive workloads. AI processing is becoming more distributed, with regional data centers hosting domain-specific models and secure edges providing low-latency, localized inferencing.
Leading data centers are exploring new approaches like offering network-as-a-service models that enable multi-tenant AI processing and extend the benefits of distributed AI across the network.
Using AI to benefit networks and operations
AI technologies, such as machine learning and data analytics, are being integrated into networks to enhance, automate, and optimize network management, performance, and security. AI enables real-time decisions, predicts network issues, adapts to changing conditions, and improves overall efficiency.
We are seeing this play out across multiple domains with new use cases optimizing performance, enhancing security, and improving efficiency. For instance, in security systems, AI improves policy and configuration management, attack prevention, and threat detection capabilities. In the radio access network (RAN), it boosts energy efficiency and resource management.
By leveraging AI/ML models, networks gain new efficiencies. These include adaptive KPI thresholding, active test workflows, and advanced root cause analysis, reducing downtime and improving user experiences.
Test and assurance assume a new role in an AI-impacted world
Testing plays a central role in addressing AI’s adoption challenges and ensuring that networks built for and enhanced with AI are robust and trustworthy.
While currently focused on validating compliance with standards and regulations, testing must evolve to support a continuous and rigorous improvement loop from design to production. This is essential for ensuring the safe evolution of networks as AI is adopted and for overcoming the challenges hindering broader AI adoption.
Emerging AI testing methodologies include realistic emulation and simulation, the use of synthetic test data, continuous testing and automation, active testing, and non-functional testing.
As a neutral and trusted partner, Spirent brings deep expertise to navigating AI’s impact on networks. With a vendor-agnostic approach and cutting-edge test and assurance capabilities, Spirent is uniquely positioned to support organizations as they confidently embrace AI-driven advancements.
Spirent eBook explores the topics discussed in this blog in more depth and includes testing recommendations and considerations for key areas where AI impacts networks, including data center interconnect fabrics, wireline IP transport networks, wireless 5G over-the-air networks, and security.
In customer engagements, Spirent is seeing various network functions being enhanced with AI. The eBook explores current practices and recommendations for testing to ensure performance for security firewall and gateway network functions, Open RAN RIC network functions, 5G core network functions, and network orchestrators.
Learn more about testing networks built for and enhanced with AI in Spirent eBook.
