The instruction could be found here.
The application is served by Heroku:
- swagger: https://sk-dominos-pizza-vouchers-api.herokuapp.com/swagger/index.html
- api-key:
034ede0baffa42e4a1186f88ff2b825b
Swashbuckle library provides the OpenAPI Specification previously known as the Swagger Specification to serve a contract of API.
The service uses the custom implementation of the API-KEY authentication in conjunction with the Role Based authorization in order to limit the access to API. It allows to handle a different set of roles based on keys.
You can use the following API key for testing purposes: 034ede0baffa42e4a1186f88ff2b825b
As the main full text search engine is used Lucene.NET. Also as an alternative it was implemented a custom engine based on Trigrams. The advantage of the Trigrams engine is simplicity when the disadvantage is twice the memory and CPU consumption compared to Lucene.NET engine.
Github Actions was chosen to build the code and to run all verifications.
Docker packs the service a container.
docker build -t dominos-pizza-vouchers-api -f Dockerfile .
docker run -it -p 5000:80 dominos-pizza-vouchers-apiHeroku serves the docker container.
heroku container:login
heroku container:push web -a dominos-pizza-vouchers-api
heroku container:release web -a dominos-pizza-vouchers-apiService: https://sk-dominos-pizza-vouchers-api.herokuapp.com/swagger/index.html
BenchmarkDotNet checks the performance of in-memory implementations: VoucherRepository, LuceneVouchersSearch and TrigramVoucherSearch. In order to get the local results you can run the following tests here.
It is very fast and does not consume memory.
| Method | N | Mean | Error | StdDev | Ratio | Gen 0 | Allocated |
|---|---|---|---|---|---|---|---|
| GetVouchers | 100 | 60.68 ns | 1.274 ns | 1.743 ns | 1.00 | 0.0153 | 128 B |
| GetVoucherById | 100 | 10.22 ns | 0.053 ns | 0.044 ns | 1.00 | - | - |
| GetVouchersByName | 100 | 112.12 ns | 2.075 ns | 1.941 ns | 1.00 | 0.0191 | 160 B |
| GetCheapestVoucherByProductCode | 100 | 16.17 ns | 0.091 ns | 0.085 ns | 1.00 | - | - |
| GetVouchers | 1000 | 195.88 ns | 3.984 ns | 6.202 ns | 1.00 | 0.0420 | 352 B |
| GetVoucherById | 1000 | 10.19 ns | 0.265 ns | 0.261 ns | 1.00 | - | - |
| GetVouchersByName | 1000 | 115.20 ns | 2.344 ns | 2.407 ns | 1.00 | 0.0191 | 160 B |
| GetCheapestVoucherByProductCode | 1000 | 16.59 ns | 0.398 ns | 0.517 ns | 1.00 | - | - |
| GetVouchers | 10000 | 185.55 ns | 3.757 ns | 4.885 ns | 1.00 | 0.0420 | 352 B |
| GetVoucherById | 10000 | 10.38 ns | 0.108 ns | 0.096 ns | 1.00 | - | - |
| GetVouchersByName | 10000 | 107.44 ns | 1.465 ns | 1.299 ns | 1.00 | 0.0191 | 160 B |
| GetCheapestVoucherByProductCode | 10000 | 16.97 ns | 0.330 ns | 0.258 ns | 1.00 | - | - |
It is very fast and consumes little memory.
| Method | N | Mean | Error | StdDev | Ratio | Gen 0 | Gen 1 | Allocated |
|---|---|---|---|---|---|---|---|---|
| Search | 1000 | 73.57 us | 1.434 us | 1.962 us | 1.00 | 19.7754 | 4.8828 | 162 KB |
| Search | 10000 | 69.75 us | 0.876 us | 0.820 us | 1.00 | 19.8975 | 3.6621 | 163 KB |
| Search | 100000 | 92.98 us | 1.803 us | 1.771 us | 1.00 | 35.1563 | 7.3242 | 288 KB |
It is very fast and consumes less memory per operation comparing to Lucene, but twice the overall memory required.
| Method | N | Mean | Error | StdDev | Ratio | Gen 0 | Gen 1 | Allocated |
|---|---|---|---|---|---|---|---|---|
| Search | 1000 | 3.566 us | 0.0253 us | 0.0211 us | 1.00 | 4.1809 | 0.4158 | 34 KB |
| Search | 10000 | 3.549 us | 0.0116 us | 0.0102 us | 1.00 | 4.1809 | 0.4158 | 34 KB |
| Search | 100000 | 3.563 us | 0.0200 us | 0.0177 us | 1.00 | 4.1809 | 0.4158 | 34 KB |
The k6 is very powerfull to run the load tests. There are 2 tests to check throughput and latency:
To measure getting all vouchers.
Script:
k6 run -u 15 -d 15s ./k6/vouchers/search.jsResult:
http_req_waiting...............: avg=457.51µs min=0s med=506.7µs max=11.35ms p(90)=1ms p(95)=1ms
http_reqs......................: 355528 23697.385807/s
To measure searching vouchers.
Script:
k6 run -u 15 -d 15s ./k6/vouchers/search.jsResult using Lucene:
http_req_waiting...............: avg=9.42ms min=513.29µs med=9.47ms max=98.39ms p(90)=13.67ms p(95)=14.75ms
http_reqs......................: 23286 1551.494041/s
Result using Trigrams:
http_req_waiting...............: avg=13.37ms min=0s med=13.16ms max=107.11ms p(90)=20.5ms p(95)=22.91ms
http_reqs......................: 16508 1099.754487/s
According to the results, Lucene 1.5x handles search queries better than Trigrams.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent