Scalable Real-Time Content Recommendation Microservice Architecture

Design a microservices-based recommendation engine for a streaming platform that can handle 10 million concurrent users. Create an architecture that uses event-driven design with Kafka for message queuing, implements machine learning recommendation algorithms with linear and non-linear collaborative filtering, and ensures sub-100ms response times. Include circuit breakers, rate limiting, and graceful degradation strategies for system resilience. Provide a detailed system design document with performance benchmarks and horizontal scaling considerations.