A staging area for in-memory computing

An in-memory staging area provides fast access to different applications. This research is based on evaluating the benefits of a distributed in-memory staging area applied to the field of Big data. With this purpose, a prototype is designed and proposed to verify the idea. Then, a working version co...

Descripción completa

Detalles Bibliográficos
Autor: Santamaria Mateu, Pol
Tipo de recurso: tesis de maestría
Fecha de publicación:2018
País:España
Institución:Universitat Oberta de Catalunya (UOC)
Repositorio:O2, repositorio institucional de la UOC
OAI Identifier:oai:openaccess.uoc.edu:10609/83667
Acceso en línea:http://hdl.handle.net/10609/83667
Access Level:acceso abierto
Palabra clave:in-memory database
big data
Alluxio
base de datos en memoria
base de dades en memòria
Big data -- TFM
Dades massives -- TFM
Datos masivos -- TFM
Descripción
Sumario:An in-memory staging area provides fast access to different applications. This research is based on evaluating the benefits of a distributed in-memory staging area applied to the field of Big data. With this purpose, a prototype is designed and proposed to verify the idea. Then, a working version comprised of the in-memory software Alluxio and the processing engine Apache Spark is deployed and evaluated. In particular, the work demonstrates the increase in performance resulting from updating the data in the in-memory staging instead of allocating space for new objects. The evaluation is conducted by running an analytic with Spark over a continuously changing dataset stored in Alluxio. The experiments reported a throughput increase of 10x when compared to storing information in a regular parallel filesystem, and an increase of 3x compared to the official deployment methodology. By updating the dataset, the Alluxio in-memory capacity stays constant at a low level compared to current deployments where its capacity decreases linearly, resulting in lower performance.