Performance Benchmark Test on Web API build with NodeJs (pure JavaScript), NestJs (TypeScript Framework), ASP.Net Core and Python Flask framework

4 simple Web Api Projects: nodeapi (using NodeJs with express server, https://docs.nestjs.com/ nestapi (using NestJs + Fastify server, https://docs.nestjs.com/) dotnetapi (using ASP.Net Core, https://docs.microsoft.com/en-us/dotnet/core/about) flaskapi The testing Api simply builds a List contains n number strings and returns the list

The Api End Points are: DotNetApi: http://localhost:5005/dotnet/{:n} NodeApi: http://localhost:5006/node/{:n} NestApi: http://localhost:5007/nest/{:n} FlaskApi: http://localhost:5008/flask/{:n}

Steps to Perform the Test

1. Clone the Project

2. Run the dotnet api

# cd to PerformanceTest folder
cd PerformanceTest
cd dotnetapi
dotnet restore
dotnet build
dotnet run

hit http://localhost:5005/dotnet/1000 on a browser to check if the api is running

3. Run the nodejs api

cd PerformanceTest
cd nodeapi
npm install
npm run start

hit http://localhost:5006/node/1000 on a browser to check if the api is running

4. Run the nestjs api

cd PerformanceTest
cd nestapi
npm install
npm run start:prod

hit http://localhost:5007/nest/1000 on a browser to check if the api is running

5. Run the Flask api

cd PerformanceTest
cd flaskapi
python3 src/app.py

hit http://localhost:5008/flask/1000 on a browser to check if the api is running

6. Run Bechmark test using wrk benchmarking tool {https://github.com/wg/wrk}

  # running 10 threads 200 connections for 30 sec, get a simple 1000 items list
  wrk -t10 -c200 -d30s http://localhost:5005/dotnet/1000
  wrk -t10 -c200 -d30s http://localhost:5006/node/1000
  wrk -t10 -c200 -d30s http://localhost:5007/nest/1000
  wrk -t10 -c200 -d30s http://localhost:5008/flask/1000
  # you can change the n to any number to simulate the length of the tasks 
  # running on the server to see how it affect the load performance 

7. Run Bechmark test using wrk hitting end point retrieve 10 items from MongoDB

  wrk -t10 -c200 -d30s http://localhost:5005/dotnet/data
  wrk -t10 -c200 -d30s http://localhost:5006/node/data

Benchmark Result without Database access

ASP.Net Core

wrk -t10 -c200 -d30s http://localhost:5005/dotnet/1000
Running 30s test @ http://localhost:5005/dotnet/1000
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     9.38ms    8.98ms 212.91ms   96.99%
    Req/Sec     2.36k   254.53     3.48k    78.37%
  704577 requests in 30.08s, 6.60GB read
  Socket errors: connect 0, read 63, write 0, timeout 0
Requests/sec:  23425.33
Transfer/sec:    224.56MB

NodeJs

wrk -t10 -c200 -d30s http://localhost:5006/node/1000  
Running 30s test @ http://localhost:5006/node/1000
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    25.98ms    3.51ms 122.33ms   92.37%
    Req/Sec   774.06     70.96     0.88k    90.27%
  231332 requests in 30.05s, 2.18GB read
  Socket errors: connect 0, read 28, write 14, timeout 0
Requests/sec:   7699.00
Transfer/sec:     74.21MB

Nest + Fastify

wrk -t10 -c200 -d30s http://localhost:5007/nest/1000
Running 30s test @ http://localhost:5007/nest/1000
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    26.25ms    3.90ms  92.63ms   87.20%
    Req/Sec   765.50     79.72     2.06k    81.10%
  228959 requests in 30.10s, 2.14GB read
Requests/sec:   7606.28
Transfer/sec:     72.83MB

Python Flask

wrk -t10 -c200 -d30s http://localhost:5008/flask/1000
Running 30s test @ http://localhost:5008/flask/1000
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   133.12ms   99.04ms   1.13s    97.89%
    Req/Sec   104.01     26.75   202.00     73.37%
  16483 requests in 30.10s, 173.56MB read
  Socket errors: connect 32, read 1063, write 31, timeout 0
Requests/sec:    547.52

Benchmark Result with simple MongoDB Database read

ASP.Net Core

wrk -t10 -c200 -d30s http://localhost:5005/dotnet/data
Running 30s test @ http://localhost:5005/dotnet/data
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    26.27ms   14.81ms 275.42ms   85.14%
    Req/Sec   798.90     99.69     1.26k    73.50%
  238731 requests in 30.04s, 751.32MB read
Requests/sec:   7947.90
Transfer/sec:     25.01MB

NodeJs

wrk -t10 -c200 -d30s http://localhost:5006/node/data
Running 30s test @ http://localhost:5006/node/data
  10 threads and 200 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   130.78ms   36.78ms 269.53ms   67.73%
    Req/Sec   153.13     30.87   242.00     67.36%
  45825 requests in 30.10s, 147.06MB read
  Socket errors: connect 0, read 58, write 0, timeout 0
Requests/sec:   1522.42
Transfer/sec:      4.89MB

Conclusion

Under the same condition and same code without Database access:

**	NodeJs	Nest	.Net Core	Python Flask
Requests Served	231332	228959	704577	16051
Req/sec	7699	7606	23425	533.22
Transfer/sec	74.21MB	72.83MB	224.56MB	5.61MB

Under the same condition and same code with Database access:

**	NodeJs	Nest	.Net Core	Python Flask
Requests Served	45825		238731
Req/sec	1522.42		7947.90
Transfer/sec	4.89MB		25.01MB

ASP.Net Core has the best performance out of 4, almost 3 times beter when no simple code, 5 time better when retrieve data from database.

Nestjs (with Fastify) has the comparible performance with Nodejs (pure JavaScript, with Express). But Nestjs provides structure to write cleaner code.

Several Considerations

1. Tested with ASP.NET Core V2.2 and V3.1. The results are similar.
2. Turn on https will reduce the performance by 40%
3. All Tests listed in this doc are test from localhost.
4. All Tests should also be performed on Cloud enviroment as well. See CloudTestReset.md