A User-Extensible Architecture for Visualization and Analysis of Time-Series Trace Data

Abstract

This thesis describes the design and implementation of Chitra95, a software system developed for the visualization and analysis of time-series trace data. Chitra95 is based upon two earlier generations of Chitra and is aimed at producing a system with broad applicability and utility in this area of research.

This thesis contributes to the area of software design for trace visualization and analysis by proposing a set of design principles towards achieving the goals of system extensibility, reusability, reliability, testability and verifiability.

These design principles are demonstrated by Chitra95, a software architecture proposed in this thesis for visualization and analysis of time-series trace data. This architecture is novel in its combination of independence from problem domain semantics; optimization for user-extensibility and code reusability; freedom from any specific user interface model; ability to simultaneously produce an integrated application and a reusable toolkit of parts that may either be customized into a turnkey system or integrated into other software systems; support for enhanced reliability, testability and verifiability; and support for an interface to the World Wide Web and for remote execution. Finally, this thesis makes the specific contribution of a data structure for representing large traces that permits the maintenance of multiple versions of a trace and retains the ability to undo modifications made to a trace.