Event Details

A Pattern-based Environment for Architectural Recovery and Evaluation

Presenter: Dr. Kamran Sartipi - Software Engineering Group, School of Computer Science, University of Waterloo
Supervisor: Dr. J. Muzio, Acting Chair, Department of Computer Science

Date: Mon, July 21, 2003
Time: 13:30:00 - 14:30:00
Place: Engineering Office Wing (EOW) Room # 430

ABSTRACT

ABSTRACT:

In a nutshell, the approaches to software architectural recovery can be classified as clustering-based techniques to allow automation in the recovery, and pattern-based techniques to incorporate domain/system knowledge and constraints in the recovery. The important issues in software architecture recovery can be viewed as: tractability of the technique in dealing with large systems; reflection of the high-level user-defined constraints in the recovery result; user involvement to control the process; and evaluation of the result. To address the above issues, we propose an environment (Alborz) for architectural recovery and evaluation based on techniques from the areas of data mining, architecture description languages, clustering, and graph pattern matching.

A data mining technique is used to extract groups of system entities with high degree of interaction as a means for defining association based similarity metrics between two system entities or two system files. Three techniques for architecture recovery are proposed.

  1. A graph pattern matching technique that is based on the user-defined conceptual architecture of the system using a query language.
  2. An incremental optimization clustering to perform hierarchical recovery by: decomposing a system into subsystems of files, which are then decomposed into modules of system entities.
  3. A partitioning technique that organizes the system files into subsystems according to a similarity threshold to control the quality of the partition.

The proposed environment provides association-based and connectivity-based metrics to evaluate the result of the recovery, and presents the recovered architecture as web pages and as graphs to be visualized.

Note: Dr. Kamran Sartipi is a candidate for a faculty position in the Department of Computer Science.