Duration: 1:14:40

123 views

3

An important class of applications is data-driven web applications, i.e., web applications that run on top of a back-end database system. Examples of such applications include online shopping sites, online auctions, and business-to-business portals. Data-driven web applications can conceptually be divided into four layers: (1) Database, which stores the persistent data, (2) Application logic, which contains the business logic for performing user actions, (3) Web site structures, which group the application logic operations into web pages and provides the navigational (hyperlinked) structure of web pages, and (4) Web site appearance, which controls the appearance of web pages, such as the background color, font size and page layout. Current development platforms use different programming models at each tier. For example, server side application development frameworks such as J2EE and Enterprise JavaBeans (EJBs) wrap relational data as Java objects and the client side usually uses a different programming model, such as AJAX or FLASH, which allows the developer to build rich clients. In this talk, I will argue that developing with different programming models at each layer is a suboptimal strategy in terms of the system development time, system maintenance and performance. I will summarize the existing platforms for developing data-driven web applications and introduce our solution for the problem, a platform based on HILDA. HILDA is a high-level language for developing data-driven web applications. It is based on UML and the relational data model and presents a unified programming model for all layers. I will show the advantage of HILDA language and discuss how Hilda run-time system can automatically partitions applications between the tiers to improve response time based on the run-time properties of applications.