d077222897
* refactor: avoid `tokio::spawn` for cache requests During the happy-path, we don't need any tokio task to drive a cache loader requests because the future issuing the request just act as a driver. If that future is cancelled, we place the cache request in an extra task. This avoid latencies due to task overhead and (task) context switches for most requests. This may remove a millisecond or two from latency but also makes the whole thing easier to analyze/profile because we don't spawn a truckload of tasks. This trick was borrowed from rskafka. * refactor: split up code |
||
|---|---|---|
| .. | ||
| benches | ||
| src | ||
| Cargo.toml | ||