頁籤選單縮合
| 題 名 | 水平擴充排程框架之設計架構--以電子發票平臺設計個案為例=Design of a Horizontally Scalable Scheduling Framework: A Case Study of an Electronic Invoice Platform |
|---|---|
| 作 者 | 蔣齊諺; 楊至中; | 書刊名 | 大數據應用與管理 |
| 卷 期 | 6 2025.12[民114.12] |
| 頁 次 | 頁20-28 |
| 分類號 | 312.49 |
| 關鍵詞 | 排程系統; 水平擴充; 系統架構設計; 軟體重構; Scheduling system; Horizontal scalability; System architecture design; Software refactoring; Spring Boot; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 本案例以電子發票核心作業流程的排程系統為 對象,針對單體架構下排程耦合度高且難以擴展的 問題,提出以多實例部署為目標的重構作法。研究 以多實例取代單機多執行緒擴充,並將核心排程抽 離,採用 Spring Boot 重寫與 Maven 多模組管理,以 降低耦合並建立可維運的邊界與依賴關係。最後透 過實作與觀察,說明該重構在提升擴展彈性與改善 瓶頸上的效果與限制,作為排程型系統現代化改造 的參考。 本研究以電子發票平台之核心排程系統為個案, 探討在單體架構下,排程作業因高度耦合與串接式 流程設計而面臨之效能瓶頸與維運困境。隨著業務 量成長,既有系統難以透過單機多執行緒方式持續 擴充,並逐漸產生排隊延遲與高度仰賴人工調整的 問題。 為因應上述限制,本案例以支援水平擴充為目 標,將排程作業自單體系統中抽離,改採多實例部 署 策 略 , 並 以 Spring Boot 重 寫 排 程 核 心 , 搭 配 Maven 多模組架構建立清楚的模組邊界與依賴治理 機制。透過此重構方式,使排程服務具備可獨立部 署、資源隔離與彈性擴充的能力,降低架構耦合與 維運複雜度。 研究採用記錄問題診斷、技術方案評估、架構設 計與實作驗證之過程,並於實際營運環境中比較重 構前後之系統表現。結果顯示,重構後排程系統於 尖峰時段之整體處理時間由原本近一小時縮短至數 分鐘內完成,排隊與人工介入情形顯著降低,系統 穩定性與維運彈性亦獲得改善。 本研究說明,在不進行完整微服務化的前提下, 透過模組化設計與多實例部署,即可有效改善排程 型系統之效能與可維運性,並為後續系統演進提供 可行且低風險之實務參考。 |
| 英文摘要 | This case study investigates the refactoring of a scheduling system within an electronic invoice platform, addressing scalability and maintenance challenges caused by a tightly coupled monolithic architecture. As business volume increased, the original design—based on sequential scheduling and single-instance execution— exhibited significant performance bottlenecks, long processing delays, and frequent reliance on manual operational intervention. To overcome these limitations, this study adopts a multi-instance deployment strategy as an alternative to single-node multi-threaded scaling. The core scheduling logic was extracted from the monolithic system and reimplemented using Spring Boot, with Maven multi- module management applied to establish clear module boundaries and dependency governance. This architectural refactoring enables independent deployment, resource isolation, and flexible horizontal scaling of scheduling services. The research follows a case study methodology, documenting problem diagnosis, technical decision- making, architectural redesign, and implementation validation. System performance before and after refactoring was evaluated in a real production environment. The results show that peak-time processing duration was reduced from nearly one hour to only a few minutes, while task queuing and manual intervention were significantly decreased. Overall system stability and operational flexibility were also improved. This study demonstrates that, without fully migrating to a microservices architecture, substantial performance and maintainability improvements can be achieved through modularization and multi-instance deployment. The findings provide a practical and low- risk reference for modernizing scheduling-oriented systems facing scalability constraints. |
本系統中英文摘要資訊取自各篇刊載內容。