Asynchronous programming lets a task start an operation that involves waiting, then makes room for other work until it completes. In Rust, async/await supports this style of control flow. It is especially useful when an application has many I/O operations in flight. section 8.2 section 1.2.2
Keep these distinctions in mind:
- Concurrency interleaves tasks; it does not require tasks to execute simultaneously. Parallelism means simultaneous execution. section 8.0
- Nonblocking I/O can use waiting time for other tasks instead of processing every request sequentially. Lightweight async tasks can support large numbers of concurrent operations. section 8.2
- Async does not remove every cost. Compare the time spent waiting with the time spent processing results before expecting a performance gain. section 8.2
Compare blocking, sequential I/O with nonblocking I/O that leaves waiting intervals available for other work. section 8.2
Choose async for the workload
An HTTP service or a client making thousands of concurrent requests is a strong candidate. A small CLI that reads one file, or a client making a few sequential requests, is usually simpler as synchronous code. Decide early: adding async later is harder than planning for it up front. section 8.8
For a concrete Rust application, an async HTTP REST service can pair request handling with database-backed create, read, update, delete, and list operations. Account for futures, an async runtime, testing, and tracing when planning the implementation; using async also introduces complexity that sequential work may not need. section 9.0 section 8.0 section 8.8
