Please design and implement a backend system for an online trading platform of Pokémon Trading Card Game.
- This online trading platform trades 4 kinds of cards only: Pikachu, Bulbasaur, Charmander, and Squirtle.
- The price of cards is between 1.00 USD and 10.00 USD.
- Users on this platform are called traders.
- There are 10K traders.
- Traders own unlimited USD and cards.
- Traders can send orders to the platform when they want to buy or sell cards at certain prices.
- A trader can only buy or sell 1 card in 1 order.
- Traders can only buy cards using USD or sell cards for USD.
- Orders are first come first serve.
- There are 2 situations to make a trade:
- When a buy order is sent to the platform, there exists an uncompleted sell order, whose price is the lowest one among all uncompleted sell orders and less than or equal to the price of the buy order. Then, a trade is made at the price of the sell order. Both buy and sell orders are completed. Otherwise, the buy order is uncompleted.
- When a sell order is sent to the platform, there exists an uncompleted buy order, whose price is the highest one among all uncompleted buy orders and greater than or equal to the price of the sell order. Then, a trade is made at the price of the buy order. Both buy and sell orders are completed. Otherwise, the sell order is uncompleted.
- Traders can view the status of their latest 50 orders.
- Traders can view the latest 50 trades on each kind of cards.
- If the sequence of orders is fixed, the results must be the same no matter how many times you execute the sequence.
- RESTful API
- Relational database (PostgreSQL, MySQL, ...)
- Containerize
- Testing
- Gracefully shutdown
- Multithreading
- Maximize performance of finishing 1M orders
- OpenAPI (Swagger)
- Set up configs using environment variables
- View logs on visualization dashboard (Kibana, Grafana, ...)
- Microservice
- Message queue (Apache Kafka, Apache Pulsar, ...)
- gRPC
- GraphQL
- Docker Compose
- Kubernetes
- Cloud computing platforms (AWS, Azure, GCP, ...)
- NoSQL
- CI/CD
- User authentication and authorization
- High availability
- ... Please commit code to your GitHub account. You must complete basic requirements within 2 weeks. You could ask for 2 more weeks to complete some advance requirements further after finishing basic requirements if you want
專案提供 Restful API 服務，使用 Go 搭配 echo 框架實現
- echo 提供高併發 HTTP 服務
- mikunalpha/goas 提供 Command as Documents，產生 Swagger API 文件
- uber-go/fx 依賴注入工具、生命週期管理，有效管理注入的組件在關閉時能夠 Gracefully shutdown
- PostgreSQL 關聯式資料庫提供 ACID Transaction 操作
- pressly/goose 資料庫版本 Migration
- NATS Streaming 提供持久化、高吞吐量的 Message Queue 服務，其實本來是使用 kafka ，但是後來想說他的 Throughput 更高就跳槽了。沒計算到之後在打算切 Partition 時遇到了阻礙
- Redis 我最初在設計分散式鎖的選擇，後來發現可以改用 Database row lock 減少程式頻繁詢問鎖的狀態
- Docker 容器化，提高服務環境的一致性
啟動本地服務
- Make 讀取 Makefile 提供新進入專案成員快速掌握服務啟動方法
- Docker Compose 編排容器、網路關聯，快速在本地啟動 NGINX load balance 服務以及其他依賴服務
專案基礎設施部署
- AWS 雲端運算服務，在 EC2 instances 上部署 Kubernetes
- Terraform IaC(infrastructure as code) 工具，協助以程式碼管理 AWS 資源
- kOps 搭配 Terraform 管理 kubernetes 叢集，提供高可用服務的基礎設施
- Helm 以套件管理模式來管理 kubernetes，整合複雜的 k8s yaml 檔案們
- PrometheusOperator 操作 Prometheus 蒐集服務時序狀態，搭配 AlertManager Slack 告警，並提供 Grafana 監控介面
- cert-manager 搭配 AWS Route53 完成 ACME 的挑戰，只要部署 Issuer 就會自動簽署、續約 TLS 憑證，提供安全的 HTTPS 服務
日誌蒐集
- Fluent Bit 輕量型的日誌蒐集系統，部署於 Kubernetes 內持續抓取並解析 container 的輸出，並導出至目標 log server
- Loki 輕量型 log 查詢系統，和 Grafana Dashboard 整合方便 trace 服務狀態
持續整合與部署
- Travis CI 跟 Github 整合提供持續整合、測試程式碼，然後他是免費ㄉ
- Github Actions 持續部署。自動建置 docker 容器、部署 Helm release，能輕鬆運用社群上編寫好的 actions，亦可輕易 fork 回來修改自己的版本，然後他也是免費ㄉ
�PTCG 交易所提供用戶建立購買卡片訂單、販售卡片訂單，並依照建立訂單的順序予以和相對應的訂單搓合！
業務邏輯共分成兩個服務 Trader 和 Matcher，Trader 提供對外 API 接口建立訂單，
訂單的撮合可以選用幾種策略 (以啟動參數選擇)：
- (建議) 在 Trader 服務中建立訂單，透過 MessageQueue 發送持久化事件，Matcher 會依序消費訂單，並維護 in memory 的二元平衡樹(RedBlack-Tree) 快速查找、新增、刪除訂單，完成交易搓合結果
- Redis 分散式鎖，搶到鎖後可以建立訂單，並在資料庫中實現搓合
- Database 分散式鎖，對一種類卡片的 row lock，搶到鎖後可以建立訂單，並在資料庫中實現搓合