María Bermejo-Corrales, Miguel A. Díaz-Corchero, Jesús Santisteban-Fernández, Juan A.Gómez Pulido
EGI Community Forum, 2015
Lugar: Bari, Italia
Fecha: 10 al 13 de noviembre de 2015
Nowadays, the rise of big data usage at companies, universities and research centers is flooding data centers with large data sets. The unexpected arrival of new large data sets can produce an undesirable behaviour in these big data infrastructures. Database saturations, nodes failures, and maintenance tasks could affect seriously users. Therefore, it is needed to create a mechanism to improve the features of current big data databases. This work presents the analysis and development of an autoscaling system which scales a distributed database according to a set of defined conditions, such as the current usage of resources. The autoscaling system of this work is carried out using a cloud computing technology to facilitate this task, in this case OpenStack, which provides an easy and powerful mechanism to manage resources through different APIs: OpenStack native API and OCCI. The database used is MongoDB, which offers scale-out storage using a method that deploys data across multiple servers (Sharding method). This work explains the analysis and design of an autoscaling system of MongoDB databases over OpenStack.