Fuzzy logic controller algorithm for placing files in a data storage system

Abstract

The problem of organizing multi-level data storage is discussed. Information loses its relevance over time and the cost of storing it on highly available media, such as solid-state drives, becomes impractical. Until now, the placement of new files in the data storage system is decided horizontally — without taking into account the multi-level organization of the system. The file migration between storage system tiers occurs over time as statistics on the frequency of requested files are accumulated. All files have metadata, such as type, size, creation date and others, from which some information about the importance of the information can be extracted and then the distribution by levels of the data storage system can be implemented at the system input. A modern data storage system represented by four levels is proposed. The first Hi-End level is intended for storing critical data with the requirements of maximum access speed and reliability. The second level, Upper Mid-Range, is intended for enterprise applications that require high access speeds. The third level, Mid-Range, is proposed to be used for organizing file storage, and the fourth, Entry Level, is proposed to be used for creating backup copies and archives. The proposed algorithm for arranging files across tiers of a data storage system takes into account metrics indicating storage requirements and selecting a level of a data storage system that meets the requirements. These metrics include availability (speed of information delivery), importance (cost of data loss due to hardware and software failures), retention period, and request frequency. Metrics are extracted from the metadata of saved files. A new solution based on the functions of a fuzzy logic controller is proposed. Its operation algorithm can be integrated into the data storage system before the process of writing a new file. The algorithm includes three main steps. At the first step, file metrics are analyzed to form the corresponding input fuzzy sets. At the second step, a logical model is used to form the final fuzzy set. At the final stage, the fuzzy output result is obtained and the file is placed at the appropriate level of the data storage system. An example of how the controller works for files with different values of metric characteristics is given. A fuzzy logic controller can be integrated into the operation of a multi-level data storage system.

Keywords

• big data
• multi-level data storage system
• data storage requirements
• file metrics
• fuzzy logic controller
• rational storage