SEMGROMI-a semantic grouping algorithm to identifying microservices using semantic similarity of user stories

Area of Science:

• Software Engineering
• Distributed Systems
• Artificial Intelligence (Natural Language Processing)

Background:

• Microservices architecture is increasingly adopted across various domains like IoT and autonomous vehicles.
• Defining microservice granularity during monolithic system migration is a critical design challenge.
• Existing methods like Domain-Driven Design (DDD) are commonly used but can be improved.

Purpose of the Study:

• To introduce SEMGROMI, a novel semantic grouping algorithm for microservice granularity definition.
• To leverage user stories and semantic similarity for identifying the number and scope of microservices.
• To optimize microservice decomposition for low coupling, high cohesion, and semantic similarity.

Main Methods:

• Utilized user stories as input for functional requirements specification.
• Employed semantic similarity analysis on textual descriptions within user stories.
• Validated SEMGROMI against DDD and a genetic algorithm in four diverse projects.

Main Results:

• SEMGROMI demonstrated high semantic cohesion and low coupling in microservice decomposition.
• The algorithm led to reduced system complexity and inter-microservice communication.
• Estimated development time was significantly decreased compared to baseline methods.

Conclusions:

• SEMGROMI offers a viable and effective approach for designing and evaluating microservices architectures.
• The semantic similarity-based technique enhances the Microservices Backlog model for design-time decisions.
• This method facilitates graphical evaluation and metric-based decision-making for microservice applications.