Managing and Securing AI Output
summary
AI workloads are container-based workloads that ingest and produce huge volumes of data and cannot tolerate latency or packet loss, so scaling requires more than bigger fabrics. Networks move from 400G → 800G switches but the real problem is the junctions between GPUs, front end and back end, where onramps, offramps and security checks create queues. F5 positions APIs and edge controls as the junction/onramp bottleneck solution to manage and secure traffic.
AI-written summary · about
related talks
10mApproach to AI at a large scale
09mAI for NI
20mAI Impact on the Backbone
10mLightning Talks Introduction - AI for NI
20mO11y in the Evolution of AI-Era Networking Infrastructure
26mAI Traffic Is Different: Rethinking Proxy Arc...
05mAI discussions and deployments in the UK
16mThe Exploding Attack Surface of Agentic AI | F5 and