F5 Powered by NVIDIA BlueField-3 DPU Accelerates AI Application Delivery

F5 recently announced the launch of BIG-IP Next for Kubernetes , a new innovative AI application delivery and application security solution designed to provide service providers and large enterprises with a centralized control point to accelerate, protect, and simplify data traffic flowing into and out of large-scale artificial intelligence ( AI ) infrastructures.

The solution leverages high-performance NVIDIA BlueField-3 DPUs to improve data center traffic efficiency, which is critical to large-scale AI deployments. With an integrated view of networking, traffic management, and security, customers will be able to maximize data center resource utilization while achieving optimal AI application performance. This not only improves infrastructure efficiency, but also enables faster and more agile AI reasoning, and can ultimately provide an enhanced AI-driven customer experience.

F5 BIG-IP Next for Kubernetes is a solution designed specifically for Kubernetes environments and has been proven in large-scale telco clouds and 5G infrastructures. With BIG-IP Next for Kubernetes, the technology now offers customized services for leading AI use cases such as inference, retrieval augmentation generation (RAG), and seamless data management and storage. Integration with NVIDIA BlueField-3 DPU minimizes hardware footprint, enables fine-grained multi-tenancy, while optimizing energy consumption and providing high-performance networking, security, and traffic management.

The integration of F5 and NVIDIA technology will allow mobile and fixed-line telecom service providers to easily transition to a cloud-native Kubernetes infrastructure to meet the growing demand for providers to adapt service functions to a cloud-native network function (CNFs) model. F5 BIG-IP Next for Kubernetes frees up CPU resources for revenue-generating applications by offloading data-intensive tasks to the BlueField-3 DPU. The solution is particularly beneficial for virtualized RAN (vRAN) or DAA for MSOs and 5G in the core network, and lays the foundation for unlocking the potential of 6G communications in the future.

Designed for high-demand service providers and large-scale infrastructures, F5 BIG-IP Next for Kubernetes delivers the following value.

1. Simplify the delivery of cloud-scale AI services: BIG-IP Next for Kubernetes seamlessly integrates with customers’ front-end networks, significantly reducing latency while providing high-performance load balancing to handle the massive data demands of AI models with hundreds of millions of parameters and up to 1 trillion operations.
2. Increased control over AI deployments: The solution provides a centralized integration point for modern AI networks with rich observability and granular information. BIG-IP Next for Kubernetes supports multiple L7 protocols beyond HTTP, ensuring enhanced inbound and outbound control at extremely high performance.
3. Protecting new AI environments: Customers can fully automate AI training and inference endpoint discovery and security functions. BIG-IP Next for Kubernetes also isolates AI applications from targeted threats , strengthens data integrity and sovereignty, and addresses encryption capabilities that are critical in modern AI environments.

“The proliferation of AI has created an unprecedented demand for advanced semiconductors and technologies. Companies are now building AI factories – highly optimized environments designed to train AI models at scale and deliver the powerful processing power required for inference at incredible speeds and with the lowest latency,” said Kunal Anand, chief technology officer and artificial intelligence officer at F5. “F5’s powerful application delivery and security services combined with NVIDIA’s full-stack accelerated computing form a powerful ecosystem. This integrated solution covers the entire AI workload stack from the hardware acceleration layer to the application interface, providing customers with enhanced observability, granular control, and performance optimization.”

“ Service providers and enterprises need accelerated computing to securely and efficiently deliver high-performance AI applications at scale in the cloud,” said Ash Bhalgat, senior director of partnerships for AI networking and security at NVIDIA . “NVIDIA is collaborating with F5 to accelerate the delivery of AI applications to better ensure peak efficiency and a seamless user experience powered by the BlueField-3 DPU. ”

“ Realizing the potential of AI requires more data processing power than the industry has ever prepared,” said Kuba Stolarski, research vice president of IDC’s Computing Systems Research Practice. “For many companies, deploying cutting-edge AI requires extensive infrastructure construction, which is often complex and expensive, making efficient and secure operations more important than ever. F5 BIG-IP Next for Kubernetes addresses the performance and security issues of large-scale AI infrastructure. By providing optimized traffic management, enterprises can achieve greater data ingestion performance and server utilization during AI model inference, which greatly improves the customer experience for users of AI applications.”

“As AI workloads explode, enterprises are seeing a surge in demand for scalable, optimized, and controlled Kubernetes traffic management,” said Todd Hathaway, global practice manager for AI, application, and API security solutions at WTT. “With F5’s BIG-IP Next for Kubernetes directly deployed on NVIDIA BlueField-3 DPUs, this proven technology now allows large-scale AI deployments at the ideal access point. WWT customers will benefit from greater data ingestion performance and GPU utilization, while getting a better user experience during inference and strategic control points for security services. With advanced technologies from F5 and NVIDIA, two of WTT’s most strategic partners, we are further strengthening our global cybersecurity mission to deliver superior digital security.”

F5 BIG-IP Next for Kubernetes with NVIDIA BlueField-3 DPU will be available in November 2024 .