New episode alert... Akamai was solving distributed infrastructure problems long before anyone was talking about AI inference. Nearly three decades later, those same architectural principles are taking on a whole new significance. In this episode of the Shared Everything podcast I talk with Akamai’s John Alexander and Ari Weil about what happens when AI inference moves beyond centralized GPU clusters and into a distributed AI grid. We dig into the hard engineering problems: where workloads run, how CPUs and GPUs are orchestrated, what latency and data gravity mean for placement, and why getting data to compute (or compute to data) is becoming one of AI’s biggest infrastructure challenges. And as they mention, a global data namespace paired with Akamai’s distributed compute infrastructure offers a compelling approach to making inference work across locations without creating a mess of regional data silos. Listen+read here... https://lnkd.in/eGHinFp7
