資訊管理學報
諶家蘭;黃敏男;
頁: 233-261
日期: 2004/07
摘要: 物件導向(Object-Oriented)與物件關聯(Object-Relational)資料庫管理系統和技術(Database Management Systems and Technology)是網際網路多媒體資料庫、工程資料庫、XML資料庫的重要和利基的整體解決方案。績效和效能評估(Performance Measurement and Evaluation)係一項主要工具去評估和衡量物件與物件關聯資料庫技術績效和效能。在常用的效能評估方法中，例如：OO1、HyperModel、OO7、BUCKY，依然承襲傳統關聯資料庫的績效評估設計，在基本結構上缺失和不足(Structural Deficiency)有其不可突破的瓶頸和缺點：(1)一般績效評估標準只針對特定應用領域，無法考應和表達使用者的特定需求和工作量模型(Workload Model)；(2)一般績效評估標準所發展的工作量產生器(Workload Generator)，因為只針對其特定領域作設計與測試，而失去了正確性和代表性。本研究以使用者需求導向(Requirements-Driven)，新創共同載具方法(Common Carrier)，以ODMG及SQL3標準為基礎，採用其學名結構(Generic Constructs)，形成物件與物件關聯資料庫工作量載具，同時進行物件分析(Object Analysis)、運算分析(Operation Analysis)、和控制分析(Control Analysis)，產生工作量特徵組和變數組(Workload Characterization)，以建立趨近泛用型之工作量模型(Toward A More Generalized Workload Model)和工作量產生器(Workload Generator)。本研究希望能以此工作量模型的設計使物件和物件關聯資料庫效能評估能具一致性、代表性、與正確性，解決目前效能評估方法在領域鎖定的困境。
關鍵字: 物件導向資料庫;物件關聯資料庫;SQL3;ODMG;資料庫績效評估;績效評估工作量模型;工作量產生器;需求分析方法;
Abstract: Object-oriented database management systems and object-relational database management systems have emerged as the niche resolution to the Internet data management, hypertext and hypermedia data management, and XML data management. The core of success of the niche technology rests upon the computing, storage, searching and retrieval capability and capacity. Database benchmark is the vital tool to measure and evaluate the ability and performance of object-oriented and object-relational database management systems. However the current and standard benchmarks are mostly relational benchmarks. Well-known object benchmark such as OO1, HyperModel, OO7, and BUCKY, are domain-specific and application-specific. When the user domain differs from that of the benchmark and the application scenario deviates from that of the benchmark, test results cannot be used or reproduced. In fact, in many cases, they mislead the users, management, and developers. These object benchmarks suffer from the structural deficiency of domain dependency and the lack of a more generic workload model. In this study, we present a requirements-driven and a generic-construct-oriented approach to tackle the research issues of workload characterization and compilation. We adopt two key industry standards of SQL3 and ODMG and adapt their object model, transaction model, and control model to create a more generic and systematic workload requirements development method from the user's perspective. From this new method, test results can be reused and reproduced. More realistic and representative workload can be generated. Furthermore, we have constructed a prototype to implement this new method and to demonstrate the feasibility and generalization of this new approach.
Keywords: Object-Oriented Database Management Systems;Object-Relational Database Management Systems;Database Benchmark;Workload Model;Benchmark Workload Generation;Performance Measurement and Evaluation;SQL3;OD;
瀏覽次數: 14057 下載次數: 80
引用 導入Endnote
頁: 233-261
日期: 2004/07
摘要: 物件導向(Object-Oriented)與物件關聯(Object-Relational)資料庫管理系統和技術(Database Management Systems and Technology)是網際網路多媒體資料庫、工程資料庫、XML資料庫的重要和利基的整體解決方案。績效和效能評估(Performance Measurement and Evaluation)係一項主要工具去評估和衡量物件與物件關聯資料庫技術績效和效能。在常用的效能評估方法中，例如：OO1、HyperModel、OO7、BUCKY，依然承襲傳統關聯資料庫的績效評估設計，在基本結構上缺失和不足(Structural Deficiency)有其不可突破的瓶頸和缺點：(1)一般績效評估標準只針對特定應用領域，無法考應和表達使用者的特定需求和工作量模型(Workload Model)；(2)一般績效評估標準所發展的工作量產生器(Workload Generator)，因為只針對其特定領域作設計與測試，而失去了正確性和代表性。本研究以使用者需求導向(Requirements-Driven)，新創共同載具方法(Common Carrier)，以ODMG及SQL3標準為基礎，採用其學名結構(Generic Constructs)，形成物件與物件關聯資料庫工作量載具，同時進行物件分析(Object Analysis)、運算分析(Operation Analysis)、和控制分析(Control Analysis)，產生工作量特徵組和變數組(Workload Characterization)，以建立趨近泛用型之工作量模型(Toward A More Generalized Workload Model)和工作量產生器(Workload Generator)。本研究希望能以此工作量模型的設計使物件和物件關聯資料庫效能評估能具一致性、代表性、與正確性，解決目前效能評估方法在領域鎖定的困境。
關鍵字: 物件導向資料庫;物件關聯資料庫;SQL3;ODMG;資料庫績效評估;績效評估工作量模型;工作量產生器;需求分析方法;
Toward a More Genernalized Workload Model for Object and Object-Relational Dbms Benchmark Method
Abstract: Object-oriented database management systems and object-relational database management systems have emerged as the niche resolution to the Internet data management, hypertext and hypermedia data management, and XML data management. The core of success of the niche technology rests upon the computing, storage, searching and retrieval capability and capacity. Database benchmark is the vital tool to measure and evaluate the ability and performance of object-oriented and object-relational database management systems. However the current and standard benchmarks are mostly relational benchmarks. Well-known object benchmark such as OO1, HyperModel, OO7, and BUCKY, are domain-specific and application-specific. When the user domain differs from that of the benchmark and the application scenario deviates from that of the benchmark, test results cannot be used or reproduced. In fact, in many cases, they mislead the users, management, and developers. These object benchmarks suffer from the structural deficiency of domain dependency and the lack of a more generic workload model. In this study, we present a requirements-driven and a generic-construct-oriented approach to tackle the research issues of workload characterization and compilation. We adopt two key industry standards of SQL3 and ODMG and adapt their object model, transaction model, and control model to create a more generic and systematic workload requirements development method from the user's perspective. From this new method, test results can be reused and reproduced. More realistic and representative workload can be generated. Furthermore, we have constructed a prototype to implement this new method and to demonstrate the feasibility and generalization of this new approach.
Keywords: Object-Oriented Database Management Systems;Object-Relational Database Management Systems;Database Benchmark;Workload Model;Benchmark Workload Generation;Performance Measurement and Evaluation;SQL3;OD;
瀏覽次數: 14057 下載次數: 80
引用 導入Endnote