Automatic Re-modularization of Clustered Codes Considering Invocation Types

Abstract

One of the approaches for software re-modularization is to transform sequential codes into distributed ones. Code clustering as a technique for software re-modularization is used on object-oriented codes. By clustering, the classes in a program are separated into several clusters, each of which contains cohesive classes. In the state of the art, the type of invocation or message passing between classes and objects is not considered in the clustering, which can impact the quality of the clusters. This paper presents an approach in which the types of invocations between classes and objects such as synchronous, one-way, or asynchronous will also be considered in the clustering. The aim is to separate asynchronous and one-way services from the clusters and move them to new clusters, automatically. So that, by running these services in parallel, extra waiting times for service invocations can be eliminated. Also, by transferring these services to new clusters, load distribution has taken place, which can reduce response time. The experiment results confirm that the proposed approach has a significant effect in reducing the overall waiting time.