What is the performance penalty when using MediatR?

Production code

I created two web API projects with identical endpoints.

• One is using MediatR to call the use cases from the application component.
• The other is using a Primary Port to expose interfaces which are implemented by the application component.

Benchmark approach

The benchmarks are calling the two APIs and measure the execution time.

I run the benchmark twice with a small difference.

1) Using the same HttpClient instance

• Instantiate the HttpClient, in the constructor of the benchmark, once for each Web API, and reuse it each time to perform the call to the web API.

Method	Mean	Error	StdDev
'Web API with MediatR - Same HttpClient'	299.2 us	1.01 us	4.09 us
'Web API with Port - Same HttpClient'	285.6 us	0.96 us	3.95 us

Result: The Web API that uses MediatR is slower by 13.6 μs = 0.014 ms

1 μs = 0.001 ms = 0.000001 s

2) Using different HttpClient instances

• Instantiate the HttpClient each time when a call to the web API is performed and dispose it immediately afterwards.

Method	Mean	Error	StdDev
'Web API with MediatR - Different HttpClient'	4.138 ms	0.0401 ms	0.0375 ms
'Web API with Port - Different HttpClient'	3.888 ms	0.0752 ms	0.1126 ms

Result: The Web API that uses MediatR is slower by 250 μs = 0.250 ms

1 ms = 0.001 s

Note

When instantiating a new HttpClient for each call, an exception is throws in some cases.

• See more details in the socket.error.md file.

Donations

If you like my work and want to support me, you can buy me a coffee: