I'm working on library what is based on okhttp3 for kotlin. Internally okhttp3 uses its own Dispatcher and it have fields, such as maxRequests and maxRequestsPerHost that determines, how many requests can we send at the moment. I want to set amount of this requests equal to amount of coroutines in Dispatchers.IO in kotlin, to have logic close to coroutines.
So, here are the questions:
How can we get the number of threads in Dispatchers.IO scope from the program code? (Maybe there is some method/public constant that i didn't find)
Alternatively we can set this number manually by Dispatchers.IO documentation. Kotlin developers write, that amount of threads "defaults to the limit of 64 threads or the number of cores (whichever is larger)". Will this information be changed or this fact is immutable?
The answer is that you don't need the IO dispatcher to work with okhttp because it's an async HTTP library. The purpose of Dispatchers.IO is making blocking operations off the main UI thread. Async operations are to be performed on the Main dispatcher because they don't block the thread.
Related
If you use a fetch method for HTTP communication in a code that does not consider asynchronous processing at all, all functions (even the main function, to take it to the extreme) that include a fetch method, even if only a little, will need to use async/await (asynchronous processing) Is it necessary to think about
Or can we limit the scope of asynchronous processing?
It is normal to go async all the way.
If you use a hexagonal / ports-and-adapters architecture, you can (sometimes) extract the I/O operations into "ports" and keep your core business logic synchronous. But the composition of the synchronous business logic and the asynchronous ports is asynchronous, so your main entry points are almost always asynchronous.
I'm writing web services in C++/CLI (not my choice) using Microsoft's Web API. A lot of functions in Web API are async, but because I'm using C++/CLI, I don't get the async/await support of C# or VB. So the fallback position is to use ContinueWith() to schedule a continuation delegate for reading the async task's result safely.
However, because C++/CLI also doesn't support inline anonymous delegates or managed lambdas, every delegate continuation must be written as a separate function somewhere. That quickly turns into spaghetti with the number of async functions in Web API.
So, to avoid the deadlock issues of Task<T>::Result, I've been trying this:
[HttpGet, Route( "get/some/dto" )]
Task< SomeDTO ^ > ^ MyActionMethod()
{
return Task::Run( gcnew Func< SomeDTO ^ >( this, &MyController::MyActionMethod2 ) );
}
SomeDTO ^ MyActionMethod2()
{
// execute code and use any task->Result calls I need without deadlocking
}
Okay, so I know this isn't great, but how bad is it? I don't yet understand enough of the guts of Web API or ASP.NET to comprehend the performance or scaling ramifications this will have.
Also, what other consequences may this have that aren't necessarily related to performance? For example, exceptions get wrapped in an extra AggregateException, which represents additional complexity and work for handling exceptions.
Your memory usage will increase with your application's parallelism. For every concurrent call to MyActionMethod you will need a separate thread with its own stack. That will cost you about 1 MB of RAM for each concurrent call. If MyActionMethod runs long enough so that 10000 instances run at once, you're looking at 10 GB of RAM. There is also CPU overhead in setting up each thread.
If concurrency is low, dropping async support won't be a problem. In that case, don't bother with Task::Run. Just change MyActionMethod to return SomeDTO^ (no Task wrapper).
Another potential concern is that lose easy use of cancellation tokens. However, for Web API it's usually fine to just let an exception propagate back to Web API, which ends up cancelling the synchronous call anyway.
Finally, if you were planning on performing any operation within your action method in parallel, you'll still need to use ContinueWith to accomplish that. Going non-async by default means you'll always perform one operation at a time. Fortunately, it's often just fine to do so.
Okay, so I know this isn't great, but how bad is it?
It's difficult to answer this without load-testing your specific scenario. But you can walk through the known semantics (taken largely from my blog).
First, when a request comes in, ASP.NET executes your handler on a thread pool thread within that request context. Your request handler calls Task.Run, which takes another thread from the thread pool and executes the actual request logic on it. The handler then returns the task returned from Task.Run; this releases the original request thread back to the thread pool.
Then, the Task.Run delegate will block on any asynchronous parts. So, this pattern has the scaling disadvantages of a regular synchronous handler, plus an extra thread context switch. Also, it uses a thread from the ASP.NET thread pool, which is not necessarily a bad thing, but in some scenarios it may throw off the ASP.NET thread pool heuristics.
Also, what other consequences may this have that aren't necessarily related to performance? For example, exceptions get wrapped in an extra AggregateException, which represents additional complexity and work for handling exceptions.
Yes, the exceptions from any .Result or Wait() calls will be wrapped in AggregateException. You may be able to avoid this by calling .GetAwaiter().GetResult() instead.
Another important consideration is that the code executing within the Task.Run is executing without a request context. So, ambient data like HttpContext.Current, current culture, thread principal, etc. are not going to be set correctly. You'll have to capture any important data before calling Task.Run and pass it down manually.
How do all the languages implements asynchronous callbacks?
For example in C++, one need to have a "monitor thread" to start a std::async. If it is started in main thread, it has to wait for the callback.
std::thread t{[]{std::async(callback_function).get();}}.detach();
v.s.
std::async(callback_function).get(); //Main thread will have to wait
What about asynchronous callbacks in JavaScript? In JS callbacks are massively used... How does V8 implement them? Does V8 create a lot of threads to listen on them and execute callback when it gets message? Or does it use one thread to listen on all the callbacks and keep refreshing?
For example,
setInterval(function(){},1000);
setInterval(function(){},2000);
Does V8 create 2 threads and monitor each callback state, or it has a pool thing to monitor all the callbacks?
V8 does not implement asynchronous functions with callbacks (including setInterval). Engine simply provides a way to execute JavaScript code.
As a V8 embedder you can create setInterval JavaScript function linked to your native C++ function that does what you want. For example, create thread or schedule some job. At this point it is your responsibility to call provided callback when it is necessary. Only one thread at a time can use V8 engine (V8 isolate instance) to execute code. This means synchronization is required if a callback needs to be called from another thread. V8 provides locking mechanism is you need this.
Another more common approach to solve this problem is to create a queue of functions for V8 to execute and use infinite queue processing loop to execute code on one thread. This is basically an event loop. This way you don't need to use execution lock, but instead use another thread to push callback function to a queue.
So it depends on a browser/Node.js/other embedder how they implement it.
TL;DR: To implement asynchronous callback is basically to allow the control flow to proceed without blocking for the callback. Before the callback function is finally called, the control flow is free to execute anything that has no dependence on the callback's result, e.g., the caller can proceed as if the callback function has returned, or the caller may yield its control to other functions.
Since the question is for general implementation rather than a specific language, my answer tries to be as general as to cover the implementation commonalities.
Different languages have different implementations for asynchronous callbacks, but the principles are the same. The key is to decouple the control flow from the code executed. They correspond to the execution context (like a thread of control with a runtime stack) and the executed task. Traditionally the execution context and the executed task are usually 1:1 associated. With asynchronous callbacks, they are decoupled.
1. The principles
To decouple the control flow from the code, it is helpful to think of every asynchronous callback as a conditional task. When the code registers an asynchronous callback, it virtually installs the task's condition in the system. The callback function is then invoked when the condition is satisfied. To support this, a condition monitoring mechanism and a task scheduler are needed, so that,
The programmer does not need to track the callback's condition;
Before the condition is satisfied, the program may proceed to execute other code that does not depend on the callback's result, without blocking on the condition;
Once the condition is satisfied, the callback is guaranteed to execute. The programmer does not need to schedule its execution;
After the callback is executed, its result is accessible to the caller.
2. Implementation for Portability
For example, if your code needs to process the data from a network connection, you do not need to write the code checking the connection state. You only registers a callback that will be invoked once the data is available for processing. The dirty work of connection checking is left to the language implementation, which is known to be tricky especially when we talk about scalability and portability.
The language implementation may employ asynchronous io, nonblocking io or a thread pool or whatever techniques to check the network state for you, and once the data is ready, the callback function is then scheduled to execute. Here the control flow of your code looks like directly going from the callback registration to the callback execution, because the language hides the intermediate steps. This is the portability story.
3. Implementation for Scalability
To hide the dirty work is only part of the whole story. The other part is that, your code itself does not need to block waiting for the task condition. It does not make sense to wait for one connection's data when you have lots of network connections simultaneously and some of them may already have data ready. The control flow of your code can simply register the callback, and then moves on with other tasks (e.g., the callbacks whose conditions have been satisfied), knowing that the registered callbacks will be executed anyway when their data are available.
If to satisfy the callback's condition does not involve much of the CPU (e.g., waiting for a timer, or waiting for the data from network), and the callback function itself is light-weighted, then single CPU (or single thread) is able to process lots of callbacks concurrently, such as incoming network requests processing. Here the control flow may look like jumping from one callback to another. This is the scalability story.
4. Implementation for Parallelism
Sometimes, the callbacks are not pending for non-blocking IO condition, but for blocking operations such as page fault; or the callbacks do not rely on any condition, but are pure computation logics. In this case, asynchronous callback does not save you the CPU waiting time (because there is no idle waiting). But since asynchronous callback implies that the callback function can be executed in parallel with the caller or other callbacks (subject to certain data sharing and synchronization constraints), the language implementation can dispatch the callback tasks to different threads, achieving the benefits of parallelism, if the platform has more than one hardware thread context. It still improves scalability.
5. Implementation for Productivity
The productivity with asynchronous callback may not be very positive when the code need to deal with chained callbacks, i.e., when callbacks register other callbacks in recursive way, known as callback hell. There are ways to rescue.
The semantics of an asynchronous callback can be explored so as to substitute the hopeless nested callbacks with other language constructs. Basically there can be two different views of callbacks:
From data flow point of view: asynchronous callback = event + task.
To register a callback essentially generates an event that will emit
when the task condition is satisfied. In this view, the chained
callbacks are just events whose processing triggers other event
emission. It can be naturally implemented in event-driven
programming, where the task execution is driven by events. Promise
and Observable may also be regarded as event-driven concept. When
multiple events are ready concurrently, their associated tasks can
be executed concurrently as well.
From control flow point of view: to register a callback yields the
control to other code, and the callback execution just resumes the
control flow once its condition is satisfied. In this view, chained
asynchronous callbacks are just resumable functions. Multiple
callbacks can be written as one after another in traditional
"synchronous" way, with yield operation in between (or await). It
actually becomes coroutine.
I haven't discussed the implementation of data passing between the asynchronous callback and its caller, but that is usually not difficult if using shared memory where caller and callback can share data. Actually Golang's channel can also be considered in line of yield/await but with its focus on data passing.
The callbacks that are passed to browser APIs, like setTimeout, are pushed into the same browser queue when the API has done its job.
The engine can check this queue when the stack is empty and push the next callback into the JS stack for execution.
You don’t have to monitor the progress of the API calls, you asked it to do a job and it will put your callback in the queue when it’s done.
Alright... I'm getting a bit confused here. The async monad allows you to use let! which will start the computation of the given async method, and suspend the thread, untill the result is available.. thats all fine, I do understand that.
Now what I dont understand is why they made an extension for the WebClient class, thats named AsyncDownloadString - Couldn't you just wrap the normal DownloadString inside an async block? I'm pretty sure, I'm missing an important point here, since I've done some testing that shows DownloadString wrapped inside an async block, still blocks the thread.
There is an important difference between the two:
The DownloadString method is synchronous - the thread that calls the method will be blocked until the whole string is downloaded (i.e. until the entire content is transferred over the internet).
On the other hand, AsyncDownloadString doesn't block the thread for a long time. It asks the operating system to start the download and then releases the thread. When the operating system receives some data, it picks a thread from the thread pool, the thread stores the data to some buffer and is again released. When all data is downloaded, the method will read all data from the buffer and resume the rest of the asynchronous workflow.
In the first case, the thread is blocked during the entire download. In the second case, threads are only busy for very short period of time (when processing received responses, but not when waiting for the server).
Internally, the AsyncDownloadString method is just a wrapper for DownloadStringAsync, so you can also find more information in the MSDN documentation.
The important point to note is that async programming is about doing operations that are not CPU bound i.e those which are IO bound. These IO bound operations are performed on IO threads (using overlapped IO feature of operating system). What this implies is that even if you wrap some factorial function inside a async block and run it inside another async block using let! binding, you won't get any benefit out of it as it will be running on CPU bound thread and the main purpose of doing async programming is to not take up a CPU bound thread when something which is of IO nature, as that CPU bound thread can be used for other purpose in the meantime the IO completes.
If you look at the various IO classes in .NET like File, Socket etc. They all have blocking as well as non blocking read and write operations. The blocking operations will wait for the IO to complete on the CPU thread and hence blocking the CPU thread till IO is done, where as the non blocking operations uses the overlapped IO API calls to perform the operation.
Async have a method to make a async block out of these non blocking APIs of Files, Socket etc. In your case calling DownloadString will block the CPU thread as it uses the blocking API of the underlying class where as AsyncDownloadString uses the non blocking - io overlapped - based API call.
the web layer is coded in asp.net with pages marked as async. Yes, the recommended way to code for aync is using the RegisterAsyncTask
I have a problem now - there are a few pages that have used AutoResetEvent and ManualResetEvent for aync and not the standard RegisterAsyncTask.
Would these objects servicing the async calls, use up the worker threads from the threadpool? (not recommended, as this would exhaust the worker threads and the server would not be able to serve other client requests
OR
would they use the IO threads? (typically IO threads are used for async calls with the RegsterAsyncTask, this is desired)
I would need to propose change to these pages based on your insights.
Any opinions please?
The reset event objects don't use different threads themselves - they just block/release the current thread based on the current state and the activities of other threads.
When you say there are other pages "that have used AutoResetEvent and ManualResetEvent for a[s]ync" what exactly do you mean? These are synchronization objects, and don't provide a way (in themselves) of making operations asynchronous. Something else must be starting a thread or using the thread pool.