Case Study: Tencent achieves secure, scalable federated learning communications with StreamNative’s Apache Pulsar

Powering Federated Learning at Tencent with Apache Pulsar

Tencent used its Angel PowerFL federated learning platform to support large-scale distributed model training while keeping data localized for privacy. The team needed a messaging system that could deliver high throughput, low latency, strong reliability, and secure data transmission across multiple parties and networks, and it evaluated several queue and synchronization options before choosing Apache Pulsar.

StreamNative’s Apache Pulsar helped Tencent build a federated communication layer with geo-replication, token authentication, topic throttling, and Kubernetes-based deployment tuning. Tencent reported that the solution has run stably in production for more than eight months and has supported Angel PowerFL for about a year, while also reducing topic-unloading issues and enabling secure synchronization across multi-party training environments.

Tencent