Optimization for the Migration Units in Information Systems based on CRUD Relations with Semi-Supervised Graph Neural Networks

Abstract

Migrating large-scale information system as a whole is impossible due to its size, thus it is necessary to divide the system into manageable migration units. Typically, optimizing the migration units can only be done by experienced engineers, and takes large amount of time. In this paper, we propose a method finding optimal migration units. To find optimal migration units, we interpret CRUD relations between functions and tables as a graph structure, and adapt semi-supervised graph neural network, Diffusion Convolutional Neural Networks, with a graph embedding method. Furthermore, we unify CRUD relation with file arrangement information and attributes of each node. In experiments, the proposed methods find practical migration units that can be applied to real examples.