Serverless & Microservices Epic Fails!

Challenges in High Scale Streaming Monitoring: • Setting up a tool to monitor streams for perceptual quality issues • Realizing the high cost and scaling bottlenecks while onboarding more streams Moving from Distributed to Monolith Architecture: • Identifying distributed systems overhead as a challenge with scaling limits and cost • Transitioning to a monolith application to simplify orchestration and reduce costs Addressing Scaling Bottlenecks: • Rearchitecting infrastructure to pack components into a single process • Migrating to a new architecture to reuse code and vertically scale detectors Deploying More Detectors for Service Scaling: • Cloning the service and parametrizing copies with different subsets of detectors • Implementing a lightweight orchestration layer for distributing customer requests Results and Takeaways: • Reducing infrastructure costs by over 90% by moving to a monolith • Improving scaling capabilities to handle thousands of streams and optimize costs Innovations with High Scale Monitoring: • Replicating a media conversion process closer to detectors for better monitoring • Monitoring all streams to enhance quality and customer experience Amélie Werner’s Leadership and Innovation: • Empowering colleagues through the redesign of user experience at Prime Video • Embracing Amazon’s Leadership Principles while prioritizing team for customer innovation