The Power of Scaling Event-Driven Microservices with Partitioned Queues

To cost-effectively implement EDA, you need to right-size for routine traffic and scale on demand with tools like Kubernetes event-driven autoscaling (KEDA). And when autoscaling, your queues need to adapt to keep delivering information in order, load balancing across many consumers. Non-exclusive queues do so, but introduce “competing consumers” so only work if the events are stateless, or if you can manage their state with a database, SAN, or cloud — which isn’t always feasible due to cost, latency or security.

Tomkins will introduce a new kind of non-exclusive queue called partitioned queues that efficiently load balance while maintaining context, order and state. You will learn about the advantages of partitioned queues, how API keys and hash functions work with them, and how they compare to Kafka consumer groups.