Technology leader modernizes DNS infrastructure with F5 to support growing AI inference throughput

Griff ShelleyProduct Marketing Manager III | F5

As organizations rapidly scale their artificial intelligence initiatives, the underlying network infrastructure faces unprecedented strain. AI workloads, particularly those focused on inference and data delivery to S3 storage endpoints, demand immense throughput, hyper-fast routing, and uncompromised security.

In an era of agents, embedded LLM’s, and the general ever-expanding footprint of AI, the Domain Name System (DNS) on which every digital interaction is instigated is a critical step in the data path that dictates whether AI initiatives can produce a return on their investment or fail.

The challenge: when legacy systems hit the ceiling

The sheer volume of requests generated by modern AI environments expose the limitations of commodity load balancing tools, challenging IT leaders to rethink how they architect for AI delivery workflows at scale. Recently, a global technology leader and hyperscaler encountered this inflection point when exploring solutions to support AI inference workflows across a footprint of over 400 global data centers.

The company needed to process high volumes of DNS requests for low-latency, inference workflows. Its legacy Citrix infrastructure became a bottleneck, with individual devices able to support only half the DNS response rates required to meet demand. This jeopardized service availability and forced the organization into highly inefficient workarounds, requiring the technology leader to deploy more hardware devices just to compensate for the legacy vendor's shortcomings.

This lack of performance also resulted in an excess of human hours spent racking, stacking, and managing the legacy devices needed to keep the infrastructure afloat. As the company prepared for high volumes of AI inference traffic, it realized the need for an infrastructure overhaul that could handle both extreme scalability and advanced traffic management, supplied by a vendor capable of matching their growth.

The solution: consolidating on F5 for unprecedented scale

To solve this problem, the firm picked F5, executing a strategic retrofit of its legacy solutions. By deploying F5 BIG-IP DNS alongside F5 BIG-IP Advanced Firewall Manager (AFM), on F5 rSeries appliances, the organization achieved the performance its workloads needed with a hardware footprint 33% smaller than the legacy solution would have required.

This technical foundation combined EDNS client subnet insertion for optimized DNS responses alongside advanced rate limiting to manage traffic surges exceeding a set safe limit. Other competitive solutions were evaluated, but could not pass the customer’s internal certification process, raising risk concerns. F5 stood out as the only partner capable of supporting the environment's scale, performance, and security demands.

The outcome: operational simplicity and AI readiness

Modernizing the footprint with F5 resolved the most pressing performance bottlenecks and significantly reduced hardware footprint. By requiring fewer systems to manage massive traffic loads, this customer reduced the operational burden of manual hardware deployments, freeing its engineering teams to focus on strategic AI expansion rather than infrastructure maintenance.

What’s next: DNS as the control layer for agentic AI

This infrastructure overhaul reflects a much broader industry reality: the fundamental shift from human-initiated web traffic to autonomous, machine-driven AI interactions. As organizations increasingly deploy agentic AI, a single human prompt can trigger a massive cascade of machine-to-machine discoveries, database queries, and API calls operating at relentless machine speed.

This shift in traffic type demands a complementary shift in addressing problems of traffic control. Before AI, we might have asked, “Which application should serve this user?” Today, that question falls more along the lines of, “Which model, region, GPU cluster, cloud, or inference provider should serve this request right now?”

The effective DNS solutions of the AI era must be ready to perform as an intelligent, high-performance control layer. And while AI traffic doesn't require a fundamentally new DNS protocol, it does dramatically amplify the need for the capabilities that make DNS effective: extreme scale, resilience, and intelligent global traffic steering.

As agentic AI systems orchestrate work across multiple models, clouds, and regions, organizations will need to dynamically steer inference requests to the most optimal, efficient, healthy, and geographically appropriate execution environment in milliseconds. As this technology leader demonstrated, ensuring true AI readiness means prioritizing a DNS and application delivery infrastructure built to handle the staggering velocity and volume of the agentic future.

Learn more about F5’s AI mission. To enhance efficiency in your own AI inference architecture, contact F5.

Also, be sure to check out more F5 customer stories to see why companies across industries choose F5.

About the Author

Griff ShelleyProduct Marketing Manager III | F5

Griff Shelley is a Product Marketing Manager at F5, specializing in hardware, software, and SaaS application delivery solutions. With a passion for connecting innovative technology to customer success, Griff drives go-to-market projects in global and local app delivery, cloud services, and AI data traffic infrastructure.

More blogs by Griff Shelley