Beyond Tokens per Second: An SLO-First Blueprint for Agentic AI Benchmarking Across Cloud Providers – Akamai Triple-T
Enterprises often compare AI inference platforms using token throughput and GPU utilization. For agentic AI, those measurements are necessary but insufficient: a system can fully utilize an expensive GPU, respond quickly and generate large numbers of tokens while still failing the underlying task, producing invalid tool calls or missing operational service-level objectives.
This session presents a reproducible, provider-neutral methodology for evaluating agentic AI inference on NVIDIA RTX PRO 6000 Blackwell Server Edition systems across cloud environments. The approach combines task success, native tool-call validity, time to first token, end-to-end latency, throughput, GPU utilization, power consumption and normalized cost into a single SLO-first measurement contract.
Attendees will learn how qualification gates, consistent workloads, infrastructure as code, immutable software and model provenance, and verified external results make cloud comparisons defensible. The session will also explain how to distinguish raw serving performance from SLO-attaining agentic throughput and how to account for provider-native infrastructure differences without overstating conclusions.
Attendees will leave with a practical benchmarking blueprint and checklist for evaluating whether an agentic inference configuration is ready for meaningful comparison across Akamai Cloud, AWS and Google Cloud.
Joseph Glover serves as Director, Field CTO for Akamai Cloud. He leads cloud technology evangelism across North America, working with customers, partners, and internal field teams to translate developments in cloud computing, edge computing, and AI into practical business and architectural strategies. His areas of focus include distributed cloud computing, multicloud portability, edge-native applications, Kubernetes, cost-effective cloud architecture, and AI inference.
Joseph has more than two decades of experience in information technology, beginning his career at AT&T in the early 2000s. Before joining Akamai, he spent more than eight years at Amazon Web Services, where he advised enterprise customers and technology partners on cloud adoption, application modernization, and scalable cloud architecture. He contributed to revenue growth and improved customer outcomes by accelerating the adoption of AWS and partner solutions and representing those technologies at industry events including HPC Summit, Cisco Live, and AWS re:Invent.
His work at Akamai spans customer advocacy, community engagement, thought leadership, field enablement, and market feedback. Joseph is particularly focused on helping enterprises build portable, resilient, and cost-effective applications across centralized cloud, distributed cloud, and edge environments while preserving architectural choice and avoiding unnecessary dependence on a single provider.
Joseph holds a Graduate Certificate in Signal and Geospatial Intelligence from The University of Texas at Dallas and a Bachelor of Science in Physics from McMurry University. His professional credentials include AWS Certified Solutions Architect – Professional and AWS Certified Machine Learning – Specialty. He has also completed the Executive Presence and Influence program through Wharton Executive Education.
Outside of work, Joseph enjoys outdoor cooking, spending time with his wife and two children, and cycling the open roads of North Texas.
