IoC Container Benchmark comparison 2016 including Microsoft.Extensions.DependencyInjection

As most of you know, my favorite is SimpleInjector. I use it in all my projects. Since Sitecore 8.2 was released yesterday (August 30th 2016), I decided to do a performance benchmark against Microsoft Extensions DependencyInjection which is used as part of Sitecore to provide DependencyInjection.

Kam has written a blog post: Dependency Injection in Sitecore 8.2 which talks about the Sitecore integration.

I ran these tests on a virtual with Intel(R) Core(TM) i7-4870HQ CPU @ 2.50GHz with 8GB memory.

My source is available on GitHub: IoC Performance

In order to do this I have taken two approaches. First, I used the repo from Daniel Palme and his blog post: IoC Container Benchmark – Performance comparison and I updated the NuGet packages. Here is a blurb from Daniels post about the setup:

The test setup

Several benchmarks are executed to test the performance of the containers in different scenarios. Every container is initialized with a couple of interfaces and their corresponding implementations. The types are registered with different lifetimes to support the various benchmarks.

Every interface is resolved 500.000 times during the benchmark and the time is measured in milliseconds. Each test is executed single threaded and multi threaded.

The benchmarks:

  • Singleton: Objects with is singleton lifetime are resolved
  • Transient: Objects with is transient lifetime are resolved
  • Combined: Objects with two dependencies (singleton and transient lifetime) are resolved
  • Complex: Objects with several nested dependencies are resolved
  • Property: Objects which require property injection are resolved
  • Generics: Objects with a generic dependency are resolved
  • IEnumerable: Several objects that implement the same interface are resolved
  • Conditional: Objects with a conditional dependency are resolved
  • Child Container: Objects are resolved trough a child container
  • Interception With Proxy: Objects with a dynamically generated wrapper are resolved
  • Prepare And Register: Initializes container and registers some basic elements (executed 3.000 times)
  • Prepare And Register And Simple Resolve: Initializes container and registers some basic elements. The resolves two objects (executed 3.000 times)

First value: Time of single-threaded execution in [ms]
Second value: Time of multi-threaded execution in [ms]

That produced the following results:

Version Singleton Transient Combined Complex Property Generics IEnumerable Conditional Child Container Interception With Proxy Prepare And Register Prepare And Register And Simple Resolve
No

 56

89

 72

104

105

124

134

134

 148

146

96

110

305

231

98

118

606

391

72

105

2  3
AutoFac 4.1.0.0  602

508

 1240

785

3760

2108

10242

6385

18682

10466

2632

1592

10363

5634

 37968

20398

Microsoft
Extensions
DependencyInjection
1.0.0.20622 156

220

140

144

320

318

729

830

198

186

738

448

18  25
Ninject 3.2.2.0  3932

2588

 13965

11564

 37197

25070

 103800

64927

93449

69829

37011

26061

91829

68581

34162

23586

62954750*

75133444*

 102117

99437

 129042  115231
SimpleInjector  3.2.2.0  66

101

87

120

 151

159

188

176

 270

218

100

131

854
520
121

144

 28926

30097

 419  3359

Second, I took the repos from:

IOCBattle by Martin From
IOCBattle by Nathanael Mann

To get the following results:

Version Singleton Reg. (ms) Singleton Resolve (1,000,000) (ms) Transient Reg. (ms) Transient Resolve (1,000,000) (ms)
new () 0.0726 57.6387 0.0683 200.6526
AutoFac Lambda 4.1.0.0 54.8336 453.2864 2.1957 6569.4292
AutoFac 4.1.0.0 4.8632 464.7712 0.7575 16183.6045
Microsoft
Extensions
DependencyInjection
1.0.0.20622 0.5311 66.0343 0.4936 198.7997
Ninject 3.2.2.0 34.5873 3343.113 1.6425 186098.403
SimpleInjector  3.2.2.0 31.0477 104.6839 2.5184 225.4272

If you have any questions or concerns, please get in touch with me. (@akshaysura13 on twitter or on Slack).

Links:
IoC Container Benchmark – Performance comparison
IoC Battle in 2015 results: Using Ninject – think again!

 

*Due to the updated AutoFac version I was getting some issues with Daniels code. I will resolve it at a later date and retest it.

one Comment

  • Richard Hauer 11 months ago

    Go again, but this time add in DryIoC, which somehow is actually faster than not using an IoC at all !?

Add your comment

Your email address will not be published.