Introduction to Async I/O

Asynchronous operations are the foundation of SPDL, and SPDL uses asyncio for orchestrating them. For advanced usage of SPDL, it is important to understand how Async I/O works, because improper usage of async I/O can slow down the whole system. However, async I/O is a programming paradigm not all Python developers are familiar with.

This section provides an introductory course on Async I/O. Instead of starting from the syntax, we start by illustrating issues we encounter when we try to build an orchestration system with traditional multi-threading.

We look at how such issues are resolved by Async I/O’s event loop, then introduce the async def and await keywords. Finally, we look at how synchronous functions can be converted to awaitable objects so that asyncio can handle sync/async functions in a unified manner.

• Implementing an Orchestration Mechanism
  • Concurrent execution and Future
  • Performing multiple tasks concurrently
  • Chaining task
  • Running multiple pipelines
  • Summary
• The Event Loop
• The async and await keywords
• Running Synchronous Functions
  • Using run_in_executor to run synchronous function in Async I/O
  • The difference of multi-threading and multi-processing
  • Using SharedMemory for faster inter-process-communication
  • Summary

See also

• Pipeline Parallelism - How SPDL uses async I/O for pipeline parallelism

• asyncio - Python’s official async I/O documentation