Automatically Refactoring Application Transactions for Microservice-oriented Architecture

Abstract

An ”application transaction”, which is a method call flow from an exported API to a data source access, is an important factor for application modernization that aims at microservice-oriented architecture. The application transaction in legacy applications is based on a traditional unified architec- ture that assumes a single application and database structure under low latency, which means that significant performance decreases can occur on microservice-oriented architectures com- posed of distributed application and database structures under high latency. Therefore, transaction refactoring is essential for application modernization. However, it is time-consuming and labor-intensive to manually identify the transaction split point in an application transaction. In this paper, we propose a method for automatically refactoring the recommendation for application transactions by identifying the split point on the basis of database operation and business contexts. We implement a framework for discovering application transactions from binary Java code and propose two methods to automatically judge whether an application transaction should be split and where the split point is located. We applied our framework and methods to actual enterprise applications, identified split points for transaction refactoring recommendation, and showed our proposal can signif- icantly reduce processing time compared with doing it manually.