Anomalies analysis and classification for web quality and reliability improvement
Jeff Tian, Nasser Alaeddine · 2009
The World Wide Web (WWW), or simply the Web, is a way of accessing information, and conducting personal and business transactions over the Internet. Initially intended as an information highway for publishing static hypertexts, the web is increasingly supporting dynamic applications, with non-trivial computations performed at run-time. With the prevalence of the web and people's reliance on it in today's society, ensuring its satisfactory reliability is also becoming increasingly important. The large scale, interconnectedness, massive user population, and high complexity of dynamic web applications, along with short development and long maintenance cycles, make it more difficult to assure high reliability for them. The increased complexity also demands disciplined methodologies and measurements for web application development and maintenance. This study addresses the web quality and reliability problems from three aspects. (1) Develop a new defect classification scheme to categorize anomalies of dynamic web applications experienced by users and observed during development and/or maintenance cycles into different defect types from both the external failure view and internal fault view. In addition, a fault-failure conversion technique based on fault exposure and severity assessment is introduced to produce the failure view of web anomalies. The analysis results are utilized to expedite the repairs of high-impact faults, which lead to cost-effective reliability improvement. (2) Develop a new defect analytic model that integrates several techniques in a hierarchical three-tiered strategy. We have selected the Unified Markov Models (UMMs) as the middle model to work under the guidance of the top-level Musa's operational profile (OP) in our overall web defect analysis framework. The bottom-level represents the critical components extracted from UMM. We developed a new dynamic ODC attributes based on Orthogonal Defect Classification (ODC) to analyze the collective defect view at the top, middle, and bottom levels of the model. This technique identifies problem areas in the web applications, find common cause, and finally suggest corrective and preventive activities to improve web reliability and software development process. (3) Integrate this study with the previous work of our research group to develop a multi-faceted quality and defect measurement for web quality assessment and improvement activities, validate the effectiveness of risk-based reliability improvement for dynamic web applications, and construct a generalized and comprehensive framework where component diversity can be evaluated for general component-based systems and system dependability can be maximized.