Jeff Tian's Research in Web-Based Systems

Web-based systems are playing a more and more important role in modern society today. Because of the pervasive nature and the massive user population, various existing software engineering approach need to be adapted for web engineering. My focus on this application area is web quality assurance (WebQA), in particular, the analysis, testing, quality/reliability improvement for such web-based applications, as described below.

Characterizing Web-based Applications for Web Quality Assurance

There are several important characteristics for web-based applications, which can be characterized from both the user and the system side:

User-focus and massive user population (user side):
Web users are the focal point of all web applications, from traditional services such as information search and retrieval, to newer applications such as e-commerce and web-based computing. The user population is extremely large and continuously growing, quite unprecedented as compared to the user population of traditional computer and software systems.
High complexity system (system side):
The pervasive nature of the web, its overall large size, distributed resources, and multi-layered structure, all contribute to the extremely high complexity of web-based systems and applications.

These factors have a strong impact on the proper choice of existing strategies or the development of new approach for web quality assurance (WebQA). A promising approach I am working on includes the following elements:

a statistical web testing strategy to ensure web reliability, or web quality from a user's perspective.
identification of problematic areas in web applications for focused WebQA.
measurement and analysis based on existing information collected by web servers to support the above WebQA activities.

Each of these elements are described in more detail below.

Web Testing Strategy: Usage-based Statistical Testing

The characteristics of the web and web-based applications lend themselves directly to statistical testing. This work is described in the following papers (you can also see the abstracts for some of these and other recent papers):

C. Kallepalli and J. Tian, "Measuring and Modeling Usage and Reliability for Statistical Web Testing", IEEE Trans. on Software Engineering, Vol.27, No.11, pp.1023-1036, Nov., 2001.
J. Tian and A. Nguyen, "Statistical Web Testing and Quality Assurance", 9th International Conference on Software Quality, pp. 263-274, Cambridge, MA, Oct., 1999.

Test-Model Construction and Suitable Model Types

Two basic usage model types for statistical web testing are: 1) flat operational profiles similar to those used in telecommunication products by Musa, and 2) Markov-chain based models. In our statistical web testing strategy, we combine both time, and integrate them with existing coverage based testing to form a three-tiered approach: 1) the top tier is a flat operational profile; 2) the middle tier is a collection of Markov chains; and 3) the bottom tier is a set of traditional testing models. The selection and usage of a lower tier model is guided by the upper tier one. Details about model construction, and suitability of specific types of models are discussed in the papers above and in the following paper (you can also see the abstracts for these papers):

J. Tian, L. Ma, Z. Li, and A.G. Koru, "A Hierarchical Strategy for Testing Web-Based Applications and Ensuring Their Reliability", 1st IEEE International Workshop on Web-Based Systems and Applications (as part of the 27th Annual International Software and Applications Conference), Dallas, Texas, Nov., 2003.
Z. Li and J. Tian, "Testing the Suitability of Markov Chains as Web Usage Models", 27th Annual International Software and Applications Conference, Dallas, Texas, Nov., 2003.

Web Workload Characterization and Reliability Analysis

Some of the issues is addressed in connection to statistical testing of web-applications above. However, because of the different operational environment, many of the existing approaches to traditional software reliability engineering need to be re-evaluated. One key measurement for reliability modeling is the workload, which is used to normalize time measurement. We have considered calendar time, as well as various other usage time, such as calendar days normalized by workload measured by counting the number of user, sessions, or individual accesses. We find that such workload measures provide a solid basis for us to evaluate the operational reliability of web applications as well as the potential for reliability growth.

Analysis of Web Problems, Risk Identification, and Quality/Reliability Improvement

We are analyzing information available in various web logs kept at the server side, to characterize the common problems, identify problematic areas for focused remedial actions, and assess performance and reliability.

This work is described in the following papers (you can also see the abstracts for some of these and other recent papers):

L. Ma and J. Tian, "Analyzing Errors and Referral Pairs to Characterize Common Problems and Improve Web Reliability", 3rd International Conference on Web Engineering, Oviedo, Spain, July, 2003.
Z. Li and J. Tian, "Analyzing Web Logs to Identify Common Errors and Improve Web Reliability", IADIS International Conference on e-Society, pp.235-242, Lisbon, Portugal, June, 2003.

Prepared by Jeff Tian (tian@engr.smu.edu). Last update July 9, 2003.

Back to Jeff Tian's home page