TokioConf

• Async

This talk explores how we identified and improved the multi-core scalability of Tokio timers. A pacing-based flow control mechanism was built on top of Tokio timers. During throughput optimization, workloads were split across multiple Tokio multi-thread runtimes instead of using a single multi-thread runtime. Surprisingly, flamegraphs revealed that splitting the runtimes reduced timer overhead, indicating a multi-core bottleneck in the timer subsystem. We formed a feedback loop: the timer imple

Player unavailable or embedding blocked? Use the YouTube link above.