Model of multi-level data storage system

Abstract

A model of multi-level organization for data storage system is studied, based on the sequential application of algorithms for vertical file distribution by the levels of data storage system, horizontal placement in sections of a certain level, and dynamic placement as a result of data migration. Selection and normalization of metadata specifying the characteristics of the stored files were performed. The model of multi-level data storage provides the storage of files in accordance with their characteristics and meets the requirements for guaranteed storage time. Representation of the storage system in the form of a matrix enables the usage of Kohonen neural network tool to arrange files by levels and sections of a specific storage system level. The application of Kohonen neural network provides the transfer from sequential execution of algorithms to placement in one step. We have proposed the model of multi-level data storage. Algorithms have been developed for file placement in a multi-level data storage system. Test examples are given which demonstrate the ability of Kohonen neural network apparatus as a tool for solving the file allocation problem in accordance with the required parameters. The combined use of file allocation algorithms gives the possibility to organize multi-level data storage in accordance with the files characteristics and the assurance of time requirements for guaranteed storage.