PRINS: scalable model inference for component-based system logs.

Area of Science:

• Software Engineering
• Model Inference
• Component-Based Systems

Background:

• Behavioral software models are crucial for software engineering but are often unavailable or outdated.
• Model inference from execution logs is a solution, but existing methods struggle with large log files.
• Scalability is a significant challenge in extracting software models from extensive system data.

Purpose of the Study:

• To address the scalability limitations of existing model inference techniques for large system logs.
• To propose a novel, scalable technique for inferring behavioral models of component-based systems.
• To evaluate the proposed technique's performance against state-of-the-art methods.

Main Methods:

• Introduced PRINS, a divide-and-conquer model inference technique.
• Inferred individual component models from corresponding execution logs.
• Merged component models by considering inter-component event flows recorded in the logs.

Main Results:

• PRINS demonstrated significantly faster processing of large logs compared to a state-of-the-art tool.
• The accuracy of models inferred by PRINS was comparable to existing methods.
• Evaluated on diverse datasets, including benchmarks and desktop application logs.

Conclusions:

• PRINS effectively overcomes the scalability issues in model inference from large execution logs.
• The technique provides accurate behavioral models for component-based systems without extra information.
• PRINS offers a practical solution for maintaining up-to-date software models in evolving systems.