Reputation: 1443
I am trying to make use of the async/await
feature of ASP.NET in my Web API project. I am not very sure whether it will make any difference in performance of my Web API service. Please find below the workflow and sample code from my application.
Work Flow:
UI Application → Web API endpoint(controller) → Call method in Web API service layer → Call another external web service. (Here we have the DB interactions, etc.)
Controller:
public async Task<IHttpActionResult> GetCountries()
{
var allCountrys = await CountryDataService.ReturnAllCountries();
if (allCountrys.Success)
{
return Ok(allCountrys.Domain);
}
return InternalServerError();
}
Service Layer:
public Task<BackOfficeResponse<List<Country>>> ReturnAllCountries()
{
var response = _service.Process<List<Country>>(BackOfficeEndpoint.CountryEndpoint, "returnCountries");
return Task.FromResult(response);
}
I tested the above code and is working. But I am not sure whether it is the correct usage of async/await
. Please share your thoughts.
Upvotes: 141
Views: 254948
Reputation: 17508
You are not leveraging async / await effectively because the request thread will be blocked while executing the synchronous method ReturnAllCountries()
The thread that is assigned to handle a request will be idly waiting while ReturnAllCountries()
does it's work.
If you can implement ReturnAllCountries()
to be asynchronous, then you would see scalability benefits. This is because the thread could be released back to the .NET thread pool to handle another request, while ReturnAllCountries()
is executing. This would allow your service to have higher throughput, by utilizing threads more efficiently.
Upvotes: 0
Reputation: 456417
I am not very sure whether it will make any difference in performance of my API.
Bear in mind that the primary benefit of asynchronous code on the server side is scalability. It won't magically make your requests run faster. I cover several "should I use async
" considerations in my article on async
ASP.NET.
I think your use case (calling other APIs) is well-suited for asynchronous code, just bear in mind that "asynchronous" does not mean "faster". The best approach is to first make your UI responsive and asynchronous; this will make your app feel faster even if it's slightly slower.
As far as the code goes, this is not asynchronous:
public Task<BackOfficeResponse<List<Country>>> ReturnAllCountries()
{
var response = _service.Process<List<Country>>(BackOfficeEndpoint.CountryEndpoint, "returnCountries");
return Task.FromResult(response);
}
You'd need a truly asynchronous implementation to get the scalability benefits of async
:
public async Task<BackOfficeResponse<List<Country>>> ReturnAllCountriesAsync()
{
return await _service.ProcessAsync<List<Country>>(BackOfficeEndpoint.CountryEndpoint, "returnCountries");
}
Or (if your logic in this method really is just a pass-through):
public Task<BackOfficeResponse<List<Country>>> ReturnAllCountriesAsync()
{
return _service.ProcessAsync<List<Country>>(BackOfficeEndpoint.CountryEndpoint, "returnCountries");
}
Note that it's easier to work from the "inside out" rather than the "outside in" like this. In other words, don't start with an asynchronous controller action and then force downstream methods to be asynchronous. Instead, identify the naturally asynchronous operations (calling external APIs, database queries, etc), and make those asynchronous at the lowest level first (Service.ProcessAsync
). Then let the async
trickle up, making your controller actions asynchronous as the last step.
And under no circumstances should you use Task.Run
in this scenario.
Upvotes: 242
Reputation: 108975
It is correct, but perhaps not useful.
As there is nothing to wait on – no calls to blocking APIs which could operate asynchronously – then you are setting up structures to track asynchronous operation (which has overhead) but then not making use of that capability.
For example, if the service layer was performing DB operations with Entity Framework which supports asynchronous calls:
public Task<BackOfficeResponse<List<Country>>> ReturnAllCountries()
{
using (db = myDBContext.Get()) {
var list = await db.Countries.Where(condition).ToListAsync();
return list;
}
}
You would allow the worker thread to do something else while the db was queried (and thus able to process another request).
Await tends to be something that needs to go all the way down: it is very hard to retro-fit into an existing system.
Upvotes: 14
Reputation: 25370
I would change your service layer to:
public Task<BackOfficeResponse<List<Country>>> ReturnAllCountries()
{
return Task.Run(() =>
{
return _service.Process<List<Country>>(BackOfficeEndpoint.CountryEndpoint, "returnCountries");
}
}
as you have it, you are still running your _service.Process
call synchronously, and gaining very little or no benefit from awaiting it.
With this approach, you are wrapping the potentially slow call in a Task
, starting it, and returning it to be awaited. Now you get the benefit of awaiting the Task
.
Upvotes: -9