DefinePK

Research Objective

To introduce an automated method for more effectively identifying potential microservices within monolithic applications by constructing a frequency-based class dependency graph and applying a clustering algorithm.

Methodology

The approach involves two stages:
1. Constructing a frequency-based class dependency network from the source code by extracting classes and methods, and then creating a weighted graph where edge weights represent the frequency of method calls.
2. Applying an iterative refinement clustering technique based on modularity optimization to group methods into potential microservices.

Methodology Flowchart

                        graph TD
    A[Start: Monolithic Source Code] --> B[Extract Classes and Methods];
    B --> C[Construct Frequency-Based Class Dependency Graph];
    C --> D[Initialize Communities];
    D --> E[Iteratively Refine Communities based on Modularity];
    E --> F[Optimize Modularity Gain];
    F --> G[Handle Singleton Nodes and Disconnected Subgraphs];
    G --> H[Convergence Achieved?];
    H -- No --> E;
    H -- Yes --> I[Identify Potential Microservices];
    I --> J[End];                    

Discussion

The proposed graph-based clustering approach effectively balances internal cohesion and external separation of microservices. It innovates by using a frequency-based weighted graph and iterative modularity optimization, leading to better cohesion, coupling, and optimal microservice detection performance. The method also addresses issues of disconnected subgraphs and singleton nodes within communities.

Key Findings

The proposed approach achieved an average F1 score of 0.88 and an average Newman-Girvan Modularity (NGM) score of 0.526, demonstrating its effectiveness in accurately identifying candidate microservices and facilitating migration from Monolithic Architecture (MA) to Microservices Architecture (MSA).

Conclusion

The developed automated method provides an effective tool for assisting developers in migrating from Monolithic Architecture to Microservices Architecture by accurately identifying potential microservices through a frequency-based class dependency graph and a modularity-based clustering algorithm.

Fact Check

- The paper reports an average F1 score of 0.88 for the proposed approach.
- The paper reports an average NGM score of 0.526 for the proposed approach.
- The paper was published in July 2024.

Mind Map