Towards Systematic Support for Utilizing Application-Level Abstraction Frameworks

点击翻译按钮获取中文摘要

Modern software development relies heavily on application-level abstraction frameworks to reduce software development overheads. For example, more than 78% of Java developers use Object-Relational Mapping frameworks to abstract database accesses; and large companies such as Amazon rely on Apache Spark to abstract distributed data analysis. Using various application-level abstraction frameworks allows developers to focus on the business logic of the application, and hence, significantly reduce the software development time and costs. ******However, using application-level abstraction frameworks can be a double-edged sword: on one hand, using these frameworks helps reduce development overheads, but on the other hand, using them incorrectly or inefficiently can result in severe software quality problems. In fact, it is common to see incorrect or suboptimal usage of abstraction frameworks due to several reasons. First, due to an incomplete understanding of the framework, developers may introduce incorrect framework usage in the application code. Second, modern application-level abstraction frameworks give developers the flexibility to customize the framework usage for different scenarios, and to tune configuration options based on real-world workloads. Not considering the scenarios and workloads would lead to sub-optimal use of the framework (e.g., having performance problems). ******The goal of this proposal is to address the above-mentioned challenges when using application-level abstraction frameworks. The proposal has three research objectives (ROs). RO1: To help developers during the software development stage, my students and I will propose techniques to automatically learn framework-specific bug patterns from software development history, and conduct user studies to understand the root causes and possible solutions of framework-related problems. RO2: To help developers during the development and operation stage, my students and I will create a lightweight real-time technique that can learn usage scenarios and typical workloads from applications that are deployed in production. The technique will provide suggestions to developers on optimizing framework-related code and configurations. RO3: To help developers during the maintenance stage, my students and I will propose techniques to automatically repair framework-specific problems by leveraging historical fixes and framework usage information in production. ******The outcome of this research will provide a complete solution to assist developers with modern framework-oriented software development. The proposed techniques will significantly improve software quality and reduce software development costs among Canadian companies. The proposed research will also train five highly qualified personnel (HQP) and allow them to contribute to state-of-the-art Software Engineering research and practice.