Static and dynamic metrics for effective object clustering

• Main Authors: Eun Sook Cho, Chul Jin Kim, Doo Dong Kim, Sung Yul Rhew
• Format: Conference Proceeding
• Language: English
• Subjects: Application software; Computer architecture; Computer displays; Electrical capacitance tomography; Hip; Internet; Java; Network servers; Telecommunication traffic; Time measurement
• DOI: 10.1109/APSEC.1998.733591

In client/server and distributed applications, the quality of object clustering plays a key role in determining the overall performance of the system. Therefore, a set of objects with higher coupling should be grouped into a single cluster so that each cluster can have a higher cohesion. As a result, the overall message traffic among objects can be greatly minimized. In addition, it should also be considered in CORBA-based applications that clusters themselves can evolve due to the dynamic object migration feature of CORBA. Hence, dynamic metrics as well as as static metrics should be developed and used in order to measure the dynamic message traffic and to tune up the system performance effectively. Various object-oriented design metrics proposed mainly deal with static coupling and cohesion, and they only consider the basic class relationships such as association, inheritance, and composition. Therefore, these metrics are not appropriate for measuring the traffic load of object messages which is closely related to the system performance. In this paper, we propose a set of metrics which considers the relevant weights on the various class relationships and estimates the static and dynamic message flow among the objects at the detailed level of member functions. By applying these metrics along with OMT or UML, we believe that clusters can be defined more efficiently and systematically, yielding high performance distributed applications.