Downloaded 14 times
Uploaded byJeremy Likness
C# Async/Await Explained
The document discusses asynchronous programming with async and await in C#. It explains that async and await were introduced to address issues with threading like overhead and blocking. Async methods use threads efficiently from the thread pool rather than creating new threads. Await yields threads rather than blocking them. The document demonstrates async programming and best practices like using Task instead of async void. It recommends using async for I/O bound work since it can improve performance over synchronous code.
C# Async/Await Explained
- 1. Your Cloud. Your Business. C#Async and Await Explained Jeremy Likness Principal Architect @JeremyLikness
- 2. Our Mission, Vision,and Values
- 3.
- 4. TODAY’S AGENDA 1. Why?Why do we need new keywords? 2. What? What exactly do async and await do? 3. How? How and when should async and await be used? 4. Q&A You have questions, I have answers
- 5.
- 6. WHY? Fundamentals • Onceupon a time, an OS was created to run apps • These apps would run in a process • Processes would be segregated into app domains • App domains would run threads Process App Domain 1 •Thread 1 •Thread 2 App Domain 2 •Thread 3 •Thread 4 • Process is the running program, i.e. the .NET CLR host • App domains provide isolation from each other and can be uniquely configured, loaded, unloaded, and secured • Threads enable management of code execution
- 7. WHY? What’s inevery thread … Thread kernel object (context) ~1KB Thread environment block (local storage data, graphics, exception handling) ~4KB – 8KB User mode stack ~1MB Kernel mode stack ~20KB
- 8. WHY? The lifeof one thread… Initialize memory Thread attach notifications sent to every DLL in the process Execute code Context Switch Execute Code Thread detach notifications Deallocate memory
- 9. WHY? Thread scheduling(1 core)
- 10. WHY? And tothink …
- 11.
- 12. WHY? A Dipin the Thread Pool • We agree threads have overhead • To address this, the CLR introduces the thread pool • Starts out empty • As tasks are dispatched, threads are created • When thread is done, it is returned to the pool and recycled • Trade-offs exist: • Less overhead (memory pressure) • Less time to allocate/spin up a thread • However, fewer threads are scheduled concurrently
- 13.
- 14. WHY? Tasks • Makeit easier to deal with threads and the thread pool • Easy to wait • Automatic ability to cancel • Simple access to result • Chainable tasks (one starts when the other finishes) • Child tasks • Parallel functions
- 15.
- 16. WHY? I/O Operations I/O Request Packet MakeI/O Request Device Driver Queue Driver Does I/O Thread Goes to Sleep Thread Wakes Up
- 17. WHY? Synchronous: TwoThreads Handle Request Blocked I/O Complete Handle Request Blocked I/O Complete
- 18. WHY? Asynchronous: OneThread Handle Request CompleteHandle Request Complete Asynchronous Asynchronous
- 19.
- 20.
- 21. What? async • Expectingto use await • Does not create new thread, always uses same thread as caller • After await may or may not use same thread (thread pool is involved, so threads are reusable) • If a SynchronizationContet exists, it will return to that thread • You can also modify this behavior using ConfigureAwait • Basically … think “yield” for threads!
- 22. What? Yield: arefresher
- 23.
- 24. Best Practices • Neverasync void (use Task instead) • Exceptions can’t be caught so they are thrown in the context (if you have one!) • Made specifically for event handlers • If you must use for event handler, try to isolate the majority of code in another await that does return a Task • Never mix async and blocking code together • Task.Wait, Task.Result are generally bad ideas • Exception is a console application • From the necessary static main, promote to an async static main with a wait • Task.Wait should become Task.When
- 25.
- 26. How? • More impactfulfor I/O bound than compute-bound • Remember the Fibonacci examples? • Check this out …
- 27.
- 28. How? • “I usuallydon’t work with multi-threading” • If you are working on the web, you are in a multi-threaded environment • If you are I/O bound, you should take advantage • Entity Framework now supports asynchronous methods! • The transformation is simple …
- 29.
- 30. How? Real WorldResults Source: http://blog.stevensanderson.com/2010/01/
- 31. How? Windows 8.x/ 10 or whatever • Windows Runtime (WinRT) • IAsyncInfo • IAsyncAction • IAsyncOperation<TResult> • IAsyncActionWithProgress<TProgress> • IAsyncOperationWithProgress<TResult, TProgress> • ThreadPool.RunAsync • IAsyncInfo.AsTask()
- 32. Recap • You arealways working with multi-threaded, don’t sell yourself short! • Compute-bound does not benefit as much from asynchronous as you might think, except to free the main context (typically your UI thread) • I/O has tremendous benefits • Async does not spin up a new thread. Instead, it establishes a state machine and makes the thread reusable and re- entrant • Await is not like Wait() because it doesn’t block and it allows you to recycle threads • If you have async I/O then USE IT! Async Task<> is your friend.
- 33.
- 34.