Backend for a real-time chat room app in a highly scalable architecture.
-
Real-time chatting using websockets.
-
Services are stateless and can be horizontally scaled.
room: creates rooms (public/protected) and handles messages.subscriber: maintains Kafka subscriber topics for each room in a Redis cluster.
-
Traefik for efficient HTTP reverse proxying and load balancing
-
gRPC for low-latency and high-throughput inter-service communication.
- with retry (Exponential backoff with jitter), timeout, and circuit breaker.
-
Graceful shutdown.
-
Observability using Prometheus + Grafana for service monitoring and OpenTelemetry + Jaeger for distributed tracing.
-
Pub/Sub using Kafka with partitioning for parallel processing.
-
Persist messages and rooms in Cassandra, A highly available and scalable NoSQL Database with tunable consistency.
-
Protect the create room API with distributed rate limiting using the Token-Bucket Algorithm with Redis.
-
Broadcasting seen, typing, joining, and leaving events to all room members.