Rethinking the Proxy Model - Implementing Envoy as a Node-Scoped Agent
CNCF [Cloud Native Computing Foundation] via YouTube
Earn a Michigan Engineering AI Certificate — Stay Ahead of the AI Revolution
Learn AI, Data Science & Business — Earn Certificates That Get You Hired
Overview
AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off Coursera Plus Annual.
Unlock All Certificates
Explore a revolutionary approach to service mesh architecture in this 32-minute conference talk that challenges the conventional sidecar proxy paradigm by implementing Envoy as a node-level agent instead of per-pod sidecars. Discover the compelling motivations behind this architectural shift and understand how moving the data plane to the node level effectively addresses common concerns associated with traditional sidecar deployments. Examine the core technical modifications required in Envoy to enable this node-scoped deployment model and learn how these changes fundamentally alter the proxy's operational characteristics. Investigate the critical aspects of configuration management in this innovative approach, including detailed explanations of how service mesh configurations are represented, applied, and managed in node-level Envoy deployments compared to the familiar sidecar model. Analyze comprehensive resource usage and performance comparisons that demonstrate the significant advantages of this consolidation approach, including reduced memory footprint, lower CPU consumption, and improved overall throughput achieved by deploying fewer, more efficient Envoy instances at the node level rather than maintaining individual proxies for each pod.
Syllabus
Rethinking the Proxy Model: Implementing Envoy as a Node-Scoped Agent - Tony Allen, Google
Taught by
CNCF [Cloud Native Computing Foundation]