Sharding high-throughput Redis without downtime

The article discusses how Inngest successfully sharded its high-throughput Redis database without any downtime. This was achieved by creating a new interface layer to coordinate interaction with the function state database, launching a state coordinator service, and implementing sharding in the production codebase. The sharding rollout strategy involved multiple stages affecting various customer segments, starting with free-tier users and ending with enterprise customers. The outcomes of this effort included 80% fewer timeouts during traffic peaks, lower CPU utilization, and a consistent latency drop from 35-50ms to 7ms.