Hello Friends!
Lately I’ve heard a lot of discussion around the new .NET Core and its performance especially on web servers.
I didn’t want to start comparing two different things, so I did patience for quite long for a more stable version.
This Monday, Microsoft announced the .NET Core version 2.0, so I feel ready to do it! Do you?
As we already mentioned, we will compare two identical things here, in terms of application, the expected response and the stability of their run times, so we will not try to put more things in the game like JSON or XMLencoders and decoders, just a simple text message. To achieve a fair comparison we will use the MVC architecture pattern on both sides, Go and .NET Core.
Prerequisites
Go (or Golang): is a rapidly growing open source programming language designed for building simple, fast, and reliable software.
There are not lot of web frameworks for Go with MVC support but, luckily for us Iris does the job.
Iris: A fast, simple and efficient micro web framework for Go. It provides a beautifully expressive and easy to use foundation for your next website, API, or distributed app.
C#: is a general-purpose, object-oriented programming language. Its development team is led by Anders Hejlsberg.
.NET Core: Develop high performance applications in less time, on any platform.
Download Go from https://golang.org/dl and .NET Core from https://www.microsoft.com/net/core.
After you’ve download and install these, you will need Iris from Go’s side. Installation is very easy, just open your terminal and execute:
go get -u github.com/kataras/iris
Benchmarking
Hardware
- Processor: Intel(R) Core(TM) i7–4710HQ CPU @ 2.50GHz 2.50GHz
- RAM: 8.00 GB
Software
- OS: Microsoft Windows [Version 10.0.15063], power plan is “High performance”
- HTTP Benchmark Tool: https://github.com/codesenberg/bombardier, latest version 1.1
- .NET Core: https://www.microsoft.com/net/core, latest version 2.0
- Iris: https://github.com/kataras/iris, latest version 8.3 built with go1.8.3
Both of the applications will just return the text “value” on request path “api/values/{id}”.
.NET Core MVC
Logo designed by Pablo Iglesias.
Created using dotnet new webapi . That webapitemplate will generate the code for you, including the return “value” on GET method requests.
Source Code
Start the .NET Core web server
$ cd netcore-mvc$ dotnet run -c ReleaseHosting environment: ProductionContent root path: C:\mygopath\src\github.com\kataras\iris\_benchmarks\netcore-mvcNow listening on: http://localhost:5000Application started. Press Ctrl+C to shut down.
Target and run the HTTP benchmark tool
$ bombardier -c 125 -n 5000000 http://localhost:5000/api/values/5Bombarding http://localhost:5000/api/values/5 with 5000000 requests using 125 connections5000000 / 5000000 [=====================================================] 100.00% 2m3sDone!Statistics Avg Stdev MaxReqs/sec 40226.03 8724.30 161919Latency 3.09ms 1.40ms 169.12msHTTP codes:1xx - 0, 2xx - 5000000, 3xx - 0, 4xx - 0, 5xx - 0others - 0Throughput: 8.91MB/s
Iris MVC
Logo designed by Santosh Anand.
Source Code
Start the Go web server
$ cd iris-mvc$ go run main.goNow listening on: http://localhost:5000Application started. Press CTRL+C to shut down.
Target and run the HTTP benchmark tool
$ bombardier -c 125 -n 5000000 http://localhost:5000/api/values/5Bombarding http://localhost:5000/api/values/5 with 5000000 requests using 125 connections5000000 / 5000000 [======================================================] 100.00% 47sDone!Statistics Avg Stdev MaxReqs/sec 105643.81 7687.79 122564Latency 1.18ms 366.55us 22.01msHTTP codes:1xx - 0, 2xx - 5000000, 3xx - 0, 4xx - 0, 5xx - 0others - 0Throughput: 19.65MB/s
For those who understand better by images, I did print my screen too!
Click here to see these screenshots.
Summary
- Time to complete the
5000000 requests- smaller is better. - Reqs/sec — bigger is better.
- Latency — smaller is better
- Throughput — bigger is better.
- Memory usage — smaller is better.
- LOC (Lines Of Code) — smaller is better.
.NET Core MVC Application, written using 86 lines of code, ran for 2 minutes and 8 seconds serving 39311.56 requests per second within 3.19ms latency in average and 229.73ms max, the memory usage of all these was ~126MB (without the dotnet host).
Iris MVC Application, written using 27 lines of code, ran for 47 seconds serving 105643.71 requests per second within 1.18ms latency in average and 22.01ms max, the memory usage of all these was ~12MB.
There is also another benchmark with templates, scroll to the bottom.
Update 20 August 2017
As Josh Clark and Scott Hanselman pointed out on this re-tweet , on .NET Core Startup.cs file the line with services.AddMvc(); can be replaced with services.AddMvcCore(); . I followed their helpful instructions and re-run the benchmarks. The article now contains the latest benchmark output for the .NET Core application with the change both Josh and Scott noted.
It had a small difference but not as huge (8.91MB/s from 8.61MB/s)
For those who want to compare with the standard services.AddMvc(); you can see the old output by pressing here.
Can you stay a bit longer for one more?
Let’s run one more benchmark, spawn 1000000 requests but this time we expect HTML generated by templates via the view engine.
.NET Core MVC with Templates
Start the .NET Core web server
$ cd netcore-mvc-templates$ dotnet run -c ReleaseHosting environment: ProductionContent root path: C:\mygopath\src\github.com\kataras\iris\_benchmarks\netcore-mvc-templatesNow listening on: http://localhost:5000Application started. Press Ctrl+C to shut down.
Target and run the HTTP benchmark tool
Bombarding http://localhost:5000 with 1000000 requests using 125 connections1000000 / 1000000 [====================================================] 100.00% 1m20sDone!Statistics Avg Stdev MaxReqs/sec 11738.60 7741.36 125887Latency 10.10ms 22.10ms 1.97sHTTP codes:1xx — 0, 2xx — 1000000, 3xx — 0, 4xx — 0, 5xx — 0others — 0Throughput: 89.03MB/s
Iris MVC with Templates
Start the Go web server
$ cd iris-mvc-templates$ go run main.goNow listening on: http://localhost:5000Application started. Press CTRL+C to shut down.
Target and run the HTTP benchmark tool
Bombarding http://localhost:5000 with 1000000 requests using 125 connections1000000 / 1000000 [======================================================] 100.00% 37sDone!Statistics Avg Stdev MaxReqs/sec 26656.76 1944.73 31188Latency 4.69ms 1.20ms 22.52msHTTP codes:1xx — 0, 2xx — 1000000, 3xx — 0, 4xx — 0, 5xx — 0others — 0Throughput: 192.51MB/s
Summary
- Time to complete the
1000000 requests— smaller is better. - Reqs/sec — bigger is better.
- Latency — smaller is better
- Memory usage — smaller is better.
- Throughput — bigger is better.
.NET Core MVC with Templates Application ran for 1 minute and 20 seconds serving 11738.60 requests per second with 89.03MB/s within 10.10ms latency in average and 1.97s max, the memory usage of all these was ~193MB (without the dotnet host).
Iris MVC with Templates Application ran for 37 seconds serving 26656.76 requests per second with 192.51MB/s within 1.18ms latency in average and 22.52ms max, the memory usage of all these was ~17MB.
What next?
Download the example source code from there and run the same benchmarks from your machine, then come back here and share your results with the rest of us!
For those who want to add other go or c# .net core web frameworks to the list please push a PR to the _benchmarks folder inside this repository.
I need to personally thanks the dev.to team for sharing my article at their twitter account, as well.
Thank you all for the 100% green feedback, have fun!
Update: Monday, 21 August 2017
A lot of people reached me saying that want to see a new benchmarking article based on the .NET Core’s lower level Kestrel this time.
So I did, follow the below link to learn the performance difference between Kestrel and Iris, it contains a sessions storage management benchmark too!
Iris Go vs .NET Core Kestrel in terms of HTTP performance_A Fair benchmark between Iris Golang and Kestrel .NET Core (C#)._hackernoon.com
I like the visual effects when I click the clap button more than once, do you? It’s simple: just click the clap button. If you feel strongly, click it more (or just hold it down).