Tiered Data Storage Model

The paper proposes an algorithm for placing data files in tiered storage in accordance with the files characteristics and the requirements for the time of their guaranteed storage. The algorithm includes the sequential application of logical rules for vertical placement of files across storage tiers, horizontal placement across segments of a certain level, and dynamic placement as a result of data migration. The placing data files is characterized by the use of file metadata analysis, and accounting for data migration allows for proactive management of storage sizes with regard to guaranteed storage times. The data storage model is formalized in the form of a matrix, the size of which equals to the number of vertical and horizontal tiers of storage. It is proposed to use the Kohonen neural network tool for placing files in tiered storage. Using Kohonen neural network allows you to move from the sequential execution of the logical rules to the placement in one step.