Consulting
A utility IoT network was losing 70-80% of its connections
About 1 million meters connected at once, and 70-80% of those connections failed, so meter data never got saved. We built a lab that reproduced production load: k6 tests across 200 pods on a 5-machine Kubernetes cluster, with OpenTelemetry metrics and Grafana dashboards. The bottlenecks sat in the network stack and the Linux kernel TCP settings. We shipped Akka.IO improvements, kernel tuning, and better load distribution.
Read the full story →