The inflexibility of monolithic architectures results in a low software systems scalability and agility, compared to the modular systems based on microservices that can be independently developed and deployed. To facilitate the migration from monolithic to microservices architecture (MSA), researchers and practitioners have proposed various approaches. However, most of these works focus on automating the identification of microservices, and few have addressed the practical aspects of generating and deploying them. This gap leaves the migration process costly and complex, relying on expert intervention and manual labour to transform monolith systems into microservices. In this paper, we suggest that LLMs constitute an exciting but ill-studied means to automate key steps of the generation of microservices, including source code restructuring, API implementation, and integration of microservice patterns. We introduce (1) a systematic approach for generating LLM prompts to guide the migration process and (2) insights drawn from applying this approach to a real-world system. Our position challenges the current reliance on manual labour for MSA migration and supports LLM-assisted automation, encouraging discussions on the feasibility, benefits and wider impact of incorporating LLMs into the migration process. We introduce (1) a systematic approach for generating LLM prompts to guide the migration process and (2) insights drawn from applying this approach to a real-world system. Our position challenges the current reliance on manual labour for MSA migration and supports LLM-assisted automation, driving the discussion on the practicality, advantages and impact of including LLMs in the migration process.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Exploring the Systematic Use of LLMs for Microservices Generation

  • Imen Trabelsi,
  • Moha Naouel,
  • Guéhéneuc Yann-Gaël

摘要

The inflexibility of monolithic architectures results in a low software systems scalability and agility, compared to the modular systems based on microservices that can be independently developed and deployed. To facilitate the migration from monolithic to microservices architecture (MSA), researchers and practitioners have proposed various approaches. However, most of these works focus on automating the identification of microservices, and few have addressed the practical aspects of generating and deploying them. This gap leaves the migration process costly and complex, relying on expert intervention and manual labour to transform monolith systems into microservices. In this paper, we suggest that LLMs constitute an exciting but ill-studied means to automate key steps of the generation of microservices, including source code restructuring, API implementation, and integration of microservice patterns. We introduce (1) a systematic approach for generating LLM prompts to guide the migration process and (2) insights drawn from applying this approach to a real-world system. Our position challenges the current reliance on manual labour for MSA migration and supports LLM-assisted automation, encouraging discussions on the feasibility, benefits and wider impact of incorporating LLMs into the migration process. We introduce (1) a systematic approach for generating LLM prompts to guide the migration process and (2) insights drawn from applying this approach to a real-world system. Our position challenges the current reliance on manual labour for MSA migration and supports LLM-assisted automation, driving the discussion on the practicality, advantages and impact of including LLMs in the migration process.