Thesis ANÁLISIS DE CONSULTAS PREDICTIVAS EN BASES DE DATOS DE SERIES DE TIEMPO
Date
2018-07
Authors
COTORAS STRAUB, SLAVKO IVO
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El presente trabajo muestra un análisis comparativo de alternativas de bases de datos de series de tiempo, orientado a usuarios que realizan consultas predictivas mediante el uso de una arquitectura de un solo nodo. El objetivo será identificar criterios al evaluar la precisión y rendimiento de los motores de bases de datos InfluxDB, Graphite y RRDtool, ejecutando un benchmark propio sobre dichos motores utilizando el mismo ambiente para su comparación.De acuerdo al análisis que se realiza en esta memoria se concluye que si se prioriza la precisión en el modelo predictivo y/o el tiempo de ejecución de las consultas, se recomienda eluso de RRDtool y Graphite. Mientras que si se busca modificar los parámetros presentes en las consultas predictivas de forma nativa, se sugieren las bases de datos InfluxDB y Graphite; siendo estas, además, buenas opciones para la gestión remota de dichas consultas.
This work shows a comparative analysis of time series database alternatives oriented to users that perform predictive queries through the use of a single node architecture. The objective will be to identify criteria when evaluating the accuracy and performance of the InfluxDB, Graphite and RRDtool database engines by running an own benchmark on them, using the same environment for comparison.According to the analysis carried out in this report, it is concluded that if the precision in the predictive model and/or the execution time of the queries is prioritized, the use of RRDtool and Graphite is recommended. While if you want to modify the parameters present in predictive queries natively, the InfluxDB and Graphite databases are suggested; being these, in addition, good options for remote management of such queries.
This work shows a comparative analysis of time series database alternatives oriented to users that perform predictive queries through the use of a single node architecture. The objective will be to identify criteria when evaluating the accuracy and performance of the InfluxDB, Graphite and RRDtool database engines by running an own benchmark on them, using the same environment for comparison.According to the analysis carried out in this report, it is concluded that if the precision in the predictive model and/or the execution time of the queries is prioritized, the use of RRDtool and Graphite is recommended. While if you want to modify the parameters present in predictive queries natively, the InfluxDB and Graphite databases are suggested; being these, in addition, good options for remote management of such queries.
Description
Catalogado desde la version PDF de la tesis.
Keywords
BASES DE DATOS , MODELOS PREDICTIVOS , SERIES DE TIEMPO