We Keep Coding

How to execute an arbitrary number of async tasks in sequential order?

I have this function:
async Task RefreshProfileInfo(List<string> listOfPlayers)
// For each player in the listOfPlayers, checks an in-memory cache if we have an entry.
// If we have a cached entry, do nothing.
// If we don't have a cached entry, fetch from backend via an API call.
This function is called very frequently, like:
await RefreshProfileInfo(playerA, playerB, playerC)
or
await RefreshProfileInfo(playerB, playerC, playerD)
or
await RefreshProfileInfo(playerE, playerF)
Ideally, if the players do not overlap each other, the calls should not affect each other (requesting PlayerE and PlayerF should not block the request for PlayerA, PlayerB, PlayerC). However, if the players DO overlap each other, the second call should wait for the first (requesting PlayerB, PlayerC, PlayerD, should wait for PlayerA, PlayerB, PlayerC to finish).
However, if that isn't possible, at the very least I'd like all calls to be sequential. (I think they should still be async, so they don't block other unrelated parts of the code).
Currently, what happens is each RefreshProfileInfo runs in parallel, which results in hitting backend every time (9 times in this example).
Instead, I want to execute them sequentially, so that only the first call hits the backend, and subsequent calls just hit cache.
What data structure/approach should I use? I'm having trouble figuring out how to "connect" the separate calls to each other. I've been playing around with Task.WhenAll() as well as SemaphoreSlim, but I can't figure out how to use them properly.
Failed attempt
The idea behind my failed attempt was to have a helper class where I could call a function, SequentialRequest(Task), and it would sequentially run all tasks invoked in this manner.
List<Task> waitingTasks = new List<Task>();
object _lock = new object();
public async Task SequentialRequest(Task func)
{
var waitingTasksCopy = new List<Task>();
lock (_lock)
{
waitingTasksCopy = new List<Task>(waitingTasks);
waitingTasks.Add(func); // Add this task to the waitingTasks (for future SequentialRequests)
}
// Wait for everything before this to finish
if (waitingTasksCopy.Count > 0)
{
await Task.WhenAll(waitingTasksCopy);
}
// Run this task
await func;
}
I thought this would work, but "func" is either run instantly (instead of waiting for earlier tasks to finish), or never run at all, depending on how I call it.
If I call it using this, it runs instantly:
async Task testTask()
{
await Task.Delay(4000);
}
If I call it using this, it never runs:
Task testTask = new Task(async () =>
{
await Task.Delay(4000);
});

Here's why your current attempt doesn't work:
// Run this task
await func;
The comment above is not describing what the code is doing. In the asynchronous world, a Task represents some operation that is already in progress. Tasks are not "run" by using await; await it a way for the current code to "asynchronously wait" for a task to complete. So no function signature taking a Task is going to work; the task is already in progress before it's even passed to that function.
Your question is actually about caching asynchronous operations. One way to do this is to cache the Task<T> itself. Currently, your cache holds the results (T); you can change your cache to hold the asynchronous operations that retrieve those results (Task<T>). For example, if your current cache type is ConcurrentDictionary<PlayerId, Player>, you could change it to ConcurrentDictionary<PlayerId, Task<Player>>.
With a cache of tasks, when your code checks for a cache entry, it will find an existing entry if the player data is loaded or has started loading. Because the Task<T> represents some asynchronous operation that is already in progress (or has already completed).
A couple of notes for this approach:
This only works for in-memory caches.
Think about how you want to handle errors. A naive cache of Task<T> will also cache error results, which is usually not desired.
The second point above is the trickier part. When an error happens, you'd probably want some additional logic to remove the errored task from the cache. Bonus points (and additional complexity) if the error handling code prevents an errored task from getting into the cache in the first place.
at the very least I'd like all calls to be sequential
Well, that's much easier. SemaphoreSlim is the asynchronous replacement for lock, so you can use a shared SemaphoreSlim. Call await mySemaphoreSlim.WaitAsync(); at the beginning of RefreshProfileInfo, put the body in a try, and in the finally block at the end of RefreshProfileInfo, call mySemaphoreSlim.Release();. That will limit all calls to RefreshProfileInfo to running sequentially.

I had the same issue in one of my projects. I had multiple threads call a single method and they all made IO calls when not found in cache. What you want to do is to add the Task to your cache and then await it. Subsequent calls will then just read the result once the task completes.
Example:
private Task RefreshProfile(Player player)
{
// cache is of type IMemoryCache
return _cache.GetOrCreate(player, entry =>
{
// expire in 30 seconds
entry.AbsoluteExpiration = DateTimeOffset.UtcNow.AddSeconds(30);
return ActualRefreshCodeThatReturnsTask(player);
});
}
Then just await in your calling code
await Task.WhenAll(RefreshProfile(Player a), RefreshProfile(Player b), RefreshProfile(Player c));

EventHub ForEach Parallel Async

Always managing to confuse myself working with async, I'm after a bit of validation/confirmation here that i'm doing what i think i'm doing in the following scenarios..
given the following trivial example:
// pretend / assume these are json msgs or something ;)
var strEvents = new List<string> { "event1", "event2", "event3" };
i can post each event to an eventhub simply as follows:
foreach (var e in strEvents)
{
// Do some things
outEventHub.Add(e); // ICollector
}
the foreach will run on a single thread, and execute each thing inside sequentially.. the posting to eventhub will also remain on the same thread too i guess??
Changing ICollector to IAsyncCollector, and achieve the following:
foreach (var e in strEvents)
{
// Do some things
await outEventHub.AddAsync(e);
}
I think i am right here in saying that the foreach will run on a single thread, the actual sending to the event hub will be pushed off elsewhere? Or at least not block that same thread..
Changing to Parallel.ForEach event as these events will be arriving 100+ or so at a time:
Parallel.ForEach(events, async (e) =>
{
// Do some things
await outEventHub.AddAsync(e);
});
Starting to get a bit hazy now, as i am not sure what really is going on now... afaik the each event has it's own thread (within the bounds of the hardware) and steps within that thread do not block it.. so this trivial example aside.
Finally, i could turn them all in to Tasks i thought..
private static async Task DoThingAsync(string e, IAsyncCollector<string> outEventHub)
{
await outEventHub.AddAsync(e);
}
var t = new List<Task>();
foreach (var e in strEvents)
{
t.Add(DoThingAsync(e, outEventHub));
}
await Task.WhenAll(t);
now i am really hazy, and i think this is prepping everything on a single thread.. and then running everything exactly at the same time, on any thread available??
I appreciate that in order to determine which is right for the job at hand benchmarking is required... but an explanation of what the framework is doing in each situation would be super helpful for me right now..

Parallel != async
This is the main idea here. Both of them have their uses, and they can be used together, but they are very different. You are mostly right with your assumptions, but let me clarify:
Simple foreach
This is non-parallel and non-async. Nothing to talk about.
Await inside foreach
This is async code that is non-parallel.
foreach (var e in strEvents)
{
// Do some things
await outEventHub.AddAsync(e);
}
This will all take place on a single thread. It takes an event, starts adding it to your event hub, and while it is being completed (I'm guessing it does some sort of network IO) it hands back the thread to the thread pool (or UI if it was called on a UI thread) so it can do other work while wating on AddAsync to return. But as you said, is is not parallel at all.
Parallel Foreach (async)
This one is a trap! In short, Parallel.Foreach is designed for synchronous workloads. We'll get back to this but first let's assume you used it with the non-async code.
Parallel foreach (sync)
A.k.a. Parallel but not async.
Parallel.ForEach(events, (e) =>
{
// Do some things
outEventHub.Add(e);
});
Each item will get its own "Task", but they won't spawn a thread. Creating threads is expensive, and in an optimal case there is no point in having more threads than CPU cores. Instead these tasks run on a ThreadPool, which has just as many Threads as optimal. Each thread takes a task, works on it, then takes another one, etc.
You can think of it as - on a 4 core machine - having 4 workers around a pile of tasks, so 4 of them are being run at a time. You can imagine that this is not ideal in case of IO bound workloads (which this most likely is). If your network is slow, you can have all 4 threads blocked on trying to send the event out, while they could be doing useful work. This leads us to...
Tasks
Async and potentially parallel (depends on the usage).
Your description is correct here, too, except for the ThreadPool, it is kikking off all the tasks at once (on the main thread), which then run on the pool's threads. While they are running, the main thread is released, which then can do other work, as needed. Up to this point it is the same as the Parallel.Foreach case. But:
What happens is that a TaskPool thread picks up a task, does the necessary preprocessing, then sends out the network request asynchronously. This means that this task will not block while waiting for the network, but rather it releases the ThreadPool thread to pick up another workitem. When the network request completes, the tasks continuation (the remaining code lines after the network request) is scheduled back to the list of tasks.
You can see that theoretically this is the most efficient process, so fast that you have to be careful not to flood your network.
Back to Parallel.Foreach and async
At this point you should be able to spot the problem. All your async lambda async (e) => { await outEventHub.AddAsync(e);} is doing is to kick off the work, it will return right after it hits the await. (Remember that async/await is releasing threads while waiting.) Parallel.Foreach returns right after it started all of them. But nothing is awaiting these tasks! These become fire and forget, which is usually a bad practice. It is like you deleted the await Task.WhenAll call from your task example.
I hope this cleared most things for you, if not, let me know what to improve on.

Why don't you send those events asynchronously in parallel, like this:
var tasks = new List<Task>();
foreach( var e in strEvents )
{
tasks.Add(outEventHub.AddAsync(e));
}
await Task.WhenAll(tasks);
await outEventHub.FlushAsync();

Using Task.Yield to overcome ThreadPool starvation while implementing producer/consumer pattern

Answering the question: Task.Yield - real usages?
I proposed to use Task.Yield allowing a pool thread to be reused by other tasks. In such pattern:
CancellationTokenSource cts;
void Start()
{
cts = new CancellationTokenSource();
// run async operation
var task = Task.Run(() => SomeWork(cts.Token), cts.Token);
// wait for completion
// after the completion handle the result/ cancellation/ errors
}
async Task<int> SomeWork(CancellationToken cancellationToken)
{
int result = 0;
bool loopAgain = true;
while (loopAgain)
{
// do something ... means a substantial work or a micro batch here - not processing a single byte
loopAgain = /* check for loop end && */ cancellationToken.IsCancellationRequested;
if (loopAgain) {
// reschedule the task to the threadpool and free this thread for other waiting tasks
await Task.Yield();
}
}
cancellationToken.ThrowIfCancellationRequested();
return result;
}
void Cancel()
{
// request cancelation
cts.Cancel();
}
But one user wrote
I don't think using Task.Yield to overcome ThreadPool starvation while
implementing producer/consumer pattern is a good idea. I suggest you
ask a separate question if you want to go into details as to why.
Anybody knows, why is not a good idea?

There are some good points left in the comments to your question. Being the user you quoted, I'd just like to sum it up: use the right tool for the job.
Using ThreadPool doesn't feel like the right tool for executing multiple continuous CPU-bound tasks, even if you try to organize some cooperative execution by turning them into state machines which yield CPU time to each other with await Task.Yield(). Thread switching is rather expensive; by doing await Task.Yield() on a tight loop you add a significant overhead. Besides, you should never take over the whole ThreadPool, as the .NET framework (and the underlying OS process) may need it for other things. On a related note, TPL even has the TaskCreationOptions.LongRunning option that requests to not run the task on a ThreadPool thread (rather, it creates a normal thread with new Thread() behind the scene).
That said, using a custom TaskScheduler with limited parallelism on some dedicated, out-of-pool threads with thread affinity for individual long-running tasks might be a different thing. At least, await continuations would be posted on the same thread, which should help reducing the switching overhead. This reminds me of a different problem I was trying to solve a while ago with ThreadAffinityTaskScheduler.
Still, depending on a particular scenario, it's usually better to use an existing well-established and tested tool. To name a few: Parallel Class, TPL Dataflow, System.Threading.Channels, Reactive Extensions.
There is also a whole range of existing industrial-strength solutions to deal with Publish-Subscribe pattern (RabbitMQ, PubNub, Redis, Azure Service Bus, Firebase Cloud Messaging (FCM), Amazon Simple Queue Service (SQS) etc).

After a bit of debating on the issue with other users - who are worried about the context switching and its influence on the performance.
I see what they are worried about.
But I meant: do something ... inside the loop to be a substantial task - usually in the form of a message handler which reads a message from the queue and processes it. The message handlers are usually user defined and the message bus executes them using some sort of dispatcher. The user can implement a handler which executes synchronously (nobody knows what the user will do), and without Task.Yield that will block the thread to process those synchronous tasks in a loop.
Not to be empty worded i added tests to github: https://github.com/BBGONE/TestThreadAffinity
They compare the ThreadAffinityTaskScheduler, .NET ThreadScheduler with BlockingCollection and .NET ThreadScheduler with Threading.Channels.
The tests show that for Ultra Short jobs the performance degradation is
around 15%. To use the Task.Yield without the performance degradation (even small) - it is not to use extremely short tasks and if the task is too short then combine shorter tasks into a bigger batch.
[The price of context switch] = [context switch duration] / ([job duration]+[context switch duration]).
In that case the influence of the switching the tasks is negligible on the performance. But it adds a better task cooperation and responsiveness of the system.
For long running tasks it is better to use a custom Scheduler which executes tasks on its own dedicated thread pool - (like the WorkStealingTaskScheduler).
For the mixed jobs - which can contain different parts - short running CPU bound, asynchronous and long running code parts. It is better to split the task into subtasks.
private async Task HandleLongRunMessage(TestMessage message, CancellationToken token = default(CancellationToken))
{
// SHORT SYNCHRONOUS TASK - execute as is on the default thread (from thread pool)
CPU_TASK(message, 50);
// IO BOUND ASYNCH TASK - used as is
await Task.Delay(50);
// BUT WRAP the LONG SYNCHRONOUS TASK inside the Task
// which is scheduled on the custom thread pool
// (to save threadpool threads)
await Task.Factory.StartNew(() => {
CPU_TASK(message, 100000);
}, token, TaskCreationOptions.DenyChildAttach, _workStealingTaskScheduler);
}

Throttling with SemaphoreSlim -- "Task.Run()" vs "new Func<Task>()"

This might not be specific to SemaphoreSlim exclusively, but basically my question is about whether there is a difference between the below two methods of throttling a collection of long running tasks, and if so, what that difference is (and when if ever to use either).
In the example below, let's say that each tracked task involves loading data from a Url (totally made up example, but is a common one that I've found for SemaphoreSlim examples).
The main difference comes down to how the individual tasks are added to the list of tracked tasks. In the first example, we call Task.Run() with a lambda, whereas in the second, we new up a Func(<Task<Result>>()) with a lambda and then immediately call that func and add the result to the tracked task list.
Examples:
Using Task.Run():
SemaphoreSlim ss = new SemaphoreSlim(_concurrentTasks);
List<string> urls = ImportUrlsFromSource();
List<Task<Result>> trackedTasks = new List<Task<Result>>();
foreach (var item in urls)
{
await ss.WaitAsync().ConfigureAwait(false);
trackedTasks.Add(Task.Run(async () =>
{
try
{
return await ProcessUrl(item);
}
catch (Exception e)
{
_log.Error($"logging some stuff");
throw;
}
finally
{
ss.Release();
}
}));
}
var results = await Task.WhenAll(trackedTasks);
Using a new Func:
SemaphoreSlim ss = new SemaphoreSlim(_concurrentTasks);
List<string> urls = ImportUrlsFromSource();
List<Task<Result>> trackedTasks = new List<Task<Result>>();
foreach (var item in urls)
{
trackedTasks.Add(new Func<Task<Result>>(async () =>
{
await ss.WaitAsync().ConfigureAwait(false);
try
{
return await ProcessUrl(item);
}
catch (Exception e)
{
_log.Error($"logging some stuff");
throw;
}
finally
{
ss.Release();
}
})());
}
var results = await Task.WhenAll(trackedTasks);

There are two differences:
Task.Run does error handling
First off all, when you call the lambda, it runs. On the other hand, Task.Run would call it. This is relevant because Task.Run does a bit of work behind the scenes. The main work it does is handling a faulted task...
If you call a lambda, and the lambda throws, it would throw before you add the Task to the list...
However, in your case, because your lambda is async, the compiler would create the Task for it (you are not making it by hand), and it will correctly handle the exception and make it available via the returned Task. Therefore this point is moot.
Task.Run prevents task attachment
Task.Run sets DenyChildAttach. This means that the tasks created inside the Task.Run run independently from (are not synchronized with) the returned Task.
For example, this code:
List<Task<int>> trackedTasks = new List<Task<int>>();
var numbers = new int[]{0, 1, 2, 3, 4};
foreach (var item in numbers)
{
trackedTasks.Add(Task.Run(async () =>
{
var x = 0;
(new Func<Task<int>>(async () =>{x = item; return x;}))().Wait();
Console.WriteLine(x);
return x;
}));
}
var results = await Task.WhenAll(trackedTasks);
Will output the numbers from 0 to 4, in unknown order. However the following code:
List<Task<int>> trackedTasks = new List<Task<int>>();
var numbers = new int[]{0, 1, 2, 3, 4};
foreach (var item in numbers)
{
trackedTasks.Add(new Func<Task<int>>(async () =>
{
var x = 0;
(new Func<Task<int>>(async () =>{x = item; return x;}))().Wait();
Console.WriteLine(x);
return x;
})());
}
var results = await Task.WhenAll(trackedTasks);
Will output the numbers from 0 to 4, in order, every time. This is odd, right? What happens is that the inner task is attached to outer one, and executed right away in the same thread. But if you use Task.Run, the inner task is not attached and scheduled independently.
This remain true even if you use await, as long as the task you await does not go to an external system...
What happens with external system? Well, for example, if your task is reading from an URL - as in your example - the system would create a TaskCompletionSource, get the Task from it, set a response handler that writes the result to the TaskCompletionSource, make the request, and return the Task. This Task is not scheduled, it running on the same thread as a parent task makes no sense. And thus, it can break the order.
Since, you are using await to wait on an external system, this point is moot too.
Conclusion
I must conclude that these are equivalent.
If you want to be safe, and make sure it works as expected, even if - in a future version - some of the above points stops being moot, then keep Task.Run. On the other hand, if you really want to optimize, use the lambda and avoid the Task.Run (very small) overhead. However, that probably won't be a bottleneck.
Addendum
When I talk about a task that goes to an external system, I refer to something that runs outside of .NET. There a bit of code that will run in .NET to interface with the external system, but the bulk of the code will not run in .NET, and thus will not be in a managed thread at all.
The consumer of the API specify nothing for this to happen. The task would be a promise task, but that is not exposed, for the consumer there is nothing special about it.
In fact, a task that goes to an external system may barely run in the CPU at all. Futhermore, it might just be waiting on something exterior to the computer (it could be the network or user input).
The pattern is as follows:
The library creates a TaskCompletionSource.
The library sets a means to recieve a notification. It can be a callback, event, message loop, hook, listening to a socket, a pipe line, waiting on a global mutex... whatever is necesary.
The library sets code to react to the notification that will call SetResult, or SetException on the TaskCompletionSource as appropiate for the notification recieved.
The library does the actual call to the external system.
The library returns TaskCompletionSource.Task.
Note: with extra care of optimization not reordering things where it should not, and with care of handling errors during the setup phase. Also, if a CancellationToken is involved, it has to be taken into account (and call SetCancelled on the TaskCompletionSource when appropiate). Also, there could be tear down necesary in the reaction to the notification (or on cancellation). Ah, do not forget to validate your parameters.
Then the external system goes and does whatever it does. Then when it finishes, or something goes wrong, gives the library the notification, and your Task is sudendtly completed, faulted... (or if cancellation happened, your Task is now cancelled) and .NET will schedule the continuations of the task as needed.
Note: async/await uses continuations behind the scenes, that is how execution resumes.
Incidentally, if you wanted to implement SempahoreSlim yourself, you would have to do something very similar to what I describe above. You can see it in my backport of SemaphoreSlim.
Let us see a couple of examples of promise tasks...
Task.Delay: when we are waiting with Task.Delay, the CPU is not spinning. This is not running in a thread. In this case the notification mechanism will be an OS timer. When the OS sees that the time of the timer has elapsed, it will call into the CLR, and then the CLR will mark the task as completed. What thread was waiting? none.
FileStream.ReadSync: when we are reading from storage with FileStream.ReadSync the actual work is done by the device. The CRL has to declare a custom event, then pass the event, the file handle and the buffer to the OS... the OS calls the device driver, the device driver interfaces with the device. As the storage device recovers the information, it will write to memory (directly on the specified buffer) via DMA technology. And when it is done, it will set an interruption, that is handled by the driver, that notifies the OS, that calls the custom event, that marks the task as completed. What thread did read the data from storage? none.
A similar pattern will be used to download from a web page, except, this time the device goes to the network. How to make an HTTP request and how the system waits for a response is beyond the scope of this answer.
It is also possible that the external system is another program, in which case it would run on a thread. But it won't be a managed thread on your process.
Your take away is that these task do not run on any of your threads. And their timing might depend on external factors. Thus, it makes no sense to think of them as running in the same thread, or that we can predict their timing (well, except of course, in the case of the timer).

Both are not very good because they create the tasks immediately. The func version is a little less overhead since it saves the Task.Run route over the thread pool just to immediately end the thread pool work and suspend on the semaphore. You don't need an async Func, you could simplify this by using an async method (possibly a local function).
But you should not do this at all. Instead, use a helper method that implements a parallel async foreach.
public static Task ForEachAsync<T>(this IEnumerable<T> source, int dop, Func<T, Task> body)
{
return Task.WhenAll(
from partition in Partitioner.Create(source).GetPartitions(dop)
select Task.Run(async delegate {
using (partition)
while (partition.MoveNext())
await body(partition.Current);
}));
}
Then you just go urls.ForEachAsync(myDop, async input => await ProcessAsync(input));
Here, the tasks are created on demand. You can even make the input stream lazy.

async await vs TaskFactory.StartNew and WaitAll

I've got a NServiceBus host that goes and downloads a whole bunch of data once a message comes through about a particular users account. One data file is about 3Mb (myob - via a webservice call) and another is about 2Mb (restful endpoint, quite fast!). To avoid waiting around for long, I've wrapped the two download calls like this:
var myobBlock = Task.Factory.StartNew(() => myobService.GetDataForUser(accountId, datablockId, CurrencyFormat.IgnoreValidator));
var account = Task.Factory.StartNew(() => accountService.DownloadMetaAccount(accountId, securityContext));
Task.WaitAll(myobBlock, account);
var myobData = myobBlock.Result;
var accountData = account.Result;
//...Process AccountData Object using myobData object
I'm wondering what the benefits are for using the new async/await patterns are here compared to the TPL-esque method I've got above. Reading Stephen Clearys notes, it seems that the above would cause the thread to sit there waiting, where as Async/Await would continue and release the thread for other work.
How would you rewrite that within the context of Async/Await and would it be beneficial? We have lots of accounts to process but its once MSMQ message per account (end of FY reporting) or per-request (ad-hoc when a customer calls up and wants their report)

The benefit of using async/await is that given a true async api (One which doesn't call sync methods over async using Task.Run and the likes, but does true async I/O work) you can avoid the allocation of any unnecessary Threads which simply waste resources only to wait on blocking I/O operations.
Lets imagine both your service methods exposed an async api, you could do the following instead of using two ThreadPool threads:
var myobBlock = myobService.GetDataForUserAsync(accountId, datablockId, CurrencyFormat.IgnoreValidator));
var account = accountService.DownloadMetaAccountAsync(accountId, securityContext));
// await till both async operations complete
await Task.WhenAll(myobBlock, account);
What will happen is that execution will yield back to the calling method until both tasks complete. When they do, continuation will resume via IOCP onto the assigned SynchronizationContext if needed.