Push Alerts to 10M Devices Under 60s
System Design diagram · made in Skemio

What it shows
- Partition user device tokens into 10k chunks across distributed Kafka topics.
- Maintain pre-warmed HTTP/2 persistent connection pools to Apple and Google APNs.
- Track delivery and failure tokens asynchronously to keep worker throughput high.
The takeaway
Fan-out speed depends on token chunking, pipelining, and pre-warmed HTTP/2 pools.
- #SystemDesign
- #Kafka
- #PushNotifications
- #Scalability
- #CloudArchitecture





