Implementação e avaliação de banco de dados chave-valor com aprendizagem de índice

Abstract

Data structures are present in many applications defining the access methods, the organization and how datasets are processed. Hash Table is a type of structure special and associative search that allows efficient searches to be carried out. what does she wear in its functioning is known as the hash or spread function. A difficulty of The Hash Table is that, depending on the distribution of the data, the modeling of the distribution function may have consequences on the performance of the structure. Furthermore, to achieve performance tolerance, some structures tend to reserve an amount of memory times greater than the size of the dataset. Currently, more adjusted models of Deep Learning (DL), which use Artificial Neural Networks, can extract complex patterns and learn the distribution of data according to its characteristics, using quantity reasonable memory. Thus, the present work aims to analyze the performance of the traditional Hash Table with the Learning Index approach, which uses DL to learn the distribution of reference data to the index of a Database (DB). How contribution, computational experiments were carried out with string data sets, in the context of Key-Value DB in memory, implementing through Hash Tables. So, With this work, the competitiveness of the Learning model of over other implementations of the Hash Table.