Engineering Perspectives

Five-nines availability allows only minutes of downtime per year. Building infrastructure across 16 mobile operators and a footprint of more than 375 million subscribers taught me that reliability isn't achieved by eliminating failure—it's achieved by designing for it.

Lessons from carrier-scale systems on blast-radius isolation, boundary architectures, circuit breakers, graceful degradation, and why operational ownership is ultimately an engineering responsibility.

Nearly three decades ago, we had to rethink internet architecture to make real-time media practical. Today, agentic AI systems are confronting surprisingly familiar problems: state, orchestration, expensive data movement, resource allocation, interoperability, and partial failure.

Drawing on my experience helping architect RealMedia and participating in the IETF standardization of RTSP, I explore five distributed-systems principles that I believe are becoming increasingly important as AI moves from prototypes to production infrastructure—and how emerging protocols such as MCP are beginning to confront many of the same architectural challenges.