As artificial intelligence (AI) applications advance, they increasingly challenge network infrastructure, particularly concerning latency and connectivity.
The rise of large-scale AI deployments brings forth new challenges, with analysts projecting that AI-related traffic will soon constitute a significant portion of overall network usage. The industry must be ready to effectively manage this expected surge. F5 is evolving its solutions to tackle the complexities of AI workloads, now incorporating real-time processing for multimodal data.
AI brings both opportunities and challenges in the realm of security. While it can enhance protective measures, it can also facilitate AI-driven cyber threats. To build AI-optimized networks, collaboration among hyperscalers, telecommunications companies, and technology firms is essential. F5 remains committed to leading advancements in this dynamic field.
Ahead of the AI & Big Data Expo Europe, Kunal Anand, Chief Technology and AI Officer at F5, discusses the company’s role and initiatives aimed at staying ahead in AI-driven networking solutions.
Industry Challenges in Latency and Connectivity
Anand: F5 has observed that AI is fundamentally reshaping application architectures. Some organizations are investing heavily in AI factories, which consist of extensive GPU clusters, while others opt for more cost-effective cloud-based solutions or small language models (SLMs).
As a result, network architectures must adapt to these evolving needs. AI factories utilize distinct networking stacks, such as InfiniBand, often coupled with specific GPUs like the H100s or NVIDIA’s forthcoming Blackwell series. Simultaneously, cloud technologies and GPU cloud offerings are progressing.
A significant trend is the concept of data gravity, where data becomes trapped in particular environments. This situation has led to the rise of multi-cloud architectures that facilitate the connection of workloads with data across different platforms, especially for retrieval-augmented generation (RAG).
As demand for RAG increases, organizations are encountering higher latency due to resource limitations, whether stemming from heavily utilized data stores or constrained GPU server availability.
Anticipating AI-Generated Traffic Challenges
Anand: F5 predicts that by the decade’s end, most applications will be driven by AI, necessitating enhancements across the network services chain. These applications will rely on APIs to interact with AI factories and third-party services, access data for RAG, and potentially expose their own APIs. Essentially, APIs will serve as the essential framework for this interconnected ecosystem.
With AI becoming central to virtually all applications, we can expect a natural uptick in AI-related traffic, which will soon dominate network utilization.
Adapting to Complex Workloads in Real Time
Anand: F5 approaches this challenge from multiple perspectives. For RAG, regardless of the data format—whether images, binary streams, or text—the retrieval method remains consistent. Customers frequently seek rapid Layer 4 load balancing, traffic management, and steering capabilities, where F5 excels. The company provides services in load balancing, traffic management, and security to ensure efficient data access for RAG, including load balancing for AI factories.
Large organizations often manage vast GPU clusters containing tens of thousands of GPUs. Given the unpredictable nature of AI workloads, these GPUs may not always be available. F5 focuses on effective traffic routing to mitigate this unpredictability.
Through these efforts, F5 enhances performance, increases throughput, and bolsters security for organizations developing AI factories and clusters.
Strengthening Network Security Against Evolving Threats
Anand: The landscape presents a myriad of AI-related security challenges. Attackers are leveraging AI to craft new payloads, exploit vulnerabilities, and initiate sophisticated attacks. For instance, tools like ChatGPT and visual transformers can compromise CAPTCHAs, particularly interactive ones, demonstrating the sophistication of these threats.
History shows that whenever attackers exploit new technologies, defenders must adapt accordingly. This often requires a reevaluation of security models, transitioning from an “allow everything, deny some” approach to one that prioritizes restriction. Many organizations are actively exploring solutions to combat AI-driven threats.
F5 is investing heavily to stay ahead of these evolving risks. Through its F5 Intelligence program, the company is developing, training, and deploying models supported by its AI Center of Excellence.
Earlier this year, F5 introduced an AI data fabric, with a dedicated team focusing on model development that benefits the entire organization, from policy creation to insights delivery. F5 is well-positioned to tackle these emerging challenges.
The Importance of Partnerships in AI Development
Anand: Collaborations are vital for the development of AI. The AI stack is intricate, encompassing various elements like electricity, data centers, hardware, servers, GPUs, memory, computational power, and networking—all of which need to function cohesively. It is rare for a single entity to manage every aspect independently.
F5 prioritizes building and maintaining the essential partnerships in computation, networking, and storage to support AI initiatives.
F5’s Commitment to Advancing AI Networking
Anand: F5 is dedicated to enhancing its technology platform. The recently launched AI Data Fabric will collaborate with the AI Center of Excellence to prepare the organization for future developments.
The company is also focusing on forming strong partnerships, with announcements forthcoming. F5 is enthusiastic about its ongoing work and the rapid global changes in the industry. Its unique position in processing worldwide traffic allows it to correlate data insights with industry trends. F5 plans to increase transparency about its research and models, including upcoming open-source contributions.