半導體測試資料STDF（Standards Test Data Format）檔案中的Data Log資料，無法如同一般半導體測試結果的等級別（Bin）資料，將它儲存進關聯式資料庫（RDBMS）後進行分析，原因為單一個測試資料STDF 檔案之中，包含的測試資料就有數百萬筆，半導體封裝測試工廠進行半導體晶圓或者積體電路的測試後，每日獲得之STDF 檔案超過數千個。如果建立傳統資料庫進行儲存將耗費大量成本且不彈性。本研究利用NoSQL Document-Oriented（MongoDB）資料庫，建構一個儲存半導體測試資料雲端資料系統，這是個將半導體測試資料STDF 轉為JSON 格式存入私有雲中的資料庫，並利用Docker Container 快速彈性的方式，進行佈署所需要的NoSQL 資料庫。讓半導體測試工程師可以方便連結到NoSQL 資料庫查詢測試結果，且公司可以在因應需要投入相對應的軟硬體資源。系統驗證部分則使用Yahoo Cloud Serving Benchmark (YCSB) 對MongoDB 進行壓力測試驗證。 The testing of the data log file in the STDF archive containing the current semiconductor test information cannot complete as Bin (levels) of chips stored in the traditional relational database for analysis, the reason being that a single STDF archive with test data contains test data with millions of record, and the STDF files acquired through semiconductor tests of chips and integrated circuits executed every day in the factory amount to several thousands. The building of a traditional database for storage is very costly and in no way elastic. This research uses the NoSQL Document-Oriented（MongoDB）database to construct a cloud data system for semiconductor chip test data, and transforms STDF into JSON format stored in a private cloud, uses Docker Container flexible and fast approach to rapid deployment of NoSQL database, with which it enables the semiconductor test engineers to make fast connections with semiconductor materials querying databases, and the company can invest in corresponding hardware and software resources. This is used to resolve the company’s problem of not being able to store the semiconductor test data log in a traditional relational database. The system is used to verify parts of Yahoo Cloud Serving Benchmark (YCSB) for stress testing to verify MongoDB.
