以關聯式資料庫與NoSQL資料庫適用時機與存取特性之探討與研究-以晶圓允收測試資料為例

Abstract

「巨量資料」是近年來很受矚目的一個議題，而資料型式不單單只有文字而已，還可以包含影音及圖像。而社群網站興起，讓資料量快速的成長，其成長速度已經遠超於目前成熟的資料庫所可處理的。以半導體產業來說，越是先進的製程，其所產生出來的資料量也是以倍數成長。與機台之間溝通的資料、晶圓測試的資料都是可以用來分析進而提升產能的要素。 其中晶圓的測試資料的分析應用上，可以影響一家半導體公司的獲利比率，於半導體產業來說，最常聽到的是「要如何提升良率」，而一個製程的良率的高低，能說明這個製程的穩定度，亦代表該製程的成熟度的重要指標之一。研發出一個全新的半導體製程過程中，必會產生出多的資料，而工程師必須從這些資料做分析與研究，修正、改良有問題或是不正確的process。 透過了解關聯式資料庫與NoSQL資料庫的特性，分析在晶圓測試資料上該如何使用這兩個不同特性的資料庫，以其適用時機與存取特性對於資訊系統的架構的相關性。

「Big Data」is a popular topic in recent years. The data type is not only for text but also includes audio and images. Because of the rise of social networking sites, it grew quickly for the quantity of data. Hence, the data growth rate cannot be handled for the existing databases. In the semiconductor industry, the advanced manufacturing process t has generated huge amount of data. The communication data between tools or wafer acceptance test (WAT) can be used to analyze and thus improved productivity. The analysis application for wafer acceptance test (WAT) that can affect a company’s profitability. It is usually to hear “How to improve yield?” in the semiconductor industry. The output quality of process yield can mean the process stability and it is an important indicator for the process maturity. To develop a new advanced process, the manufacturing process will output more data and engineers must consider those data to analysis and research to improve yield or incorrect process follows. This study analyzes two different database systems: relational database and NoSQL database features. This study also analyzes wafer acceptance test (WAT) data that based on the two different databases. We also discuss the usage timing and the characteristic of data access for information systems. Discussion and conclusions are also included in the final section.