Detalhes bibliográficos
| Ano de defesa: |
2017 |
| Autor(a) principal: |
Fonseca, Fábio Miguel Blasak da
 |
| Orientador(a): |
De Rose, César Augusto Fonticielha
|
| Banca de defesa: |
Não Informado pela instituição |
| Tipo de documento: |
Dissertação
|
| Tipo de acesso: |
Acesso aberto |
| Idioma: |
por |
| Instituição de defesa: |
Pontifícia Universidade Católica do Rio Grande do Sul
|
| Programa de Pós-Graduação: |
Programa de Pós-Graduação em Ciência da Computação
|
| Departamento: |
Escola Politécnica
|
| País: |
Brasil
|
| Palavras-chave em Português: |
|
| Área do conhecimento CNPq: |
|
| Link de acesso: |
http://tede2.pucrs.br/tede2/handle/tede/7945
|
Resumo: |
The growing need to extend IT (Information Technology) resources to meet business needs has raised concerns about how to increase capacity with lower cost and greater use of data center. Therefore, in order to avoid underutilization of infrastructure resources virtualization is a trend towards cost reduction and consolidation of the server infrastructure, thus taking advantage of existing assets. However, with virtualization growth, there is a problem related to resources concurrence in consolidated environments, where diskintensive applications such as databases can be impacted in this type of environment, if they do not have their resources managed properly, can generate performance degradation and increasing execution time respectively. In order to optimize performance and reduce I/O contention, Kassiano J.M. [19] presented a study on the acceleration of Hadoop applications through manual adjustment of disk resource allocation, showing that it is possible to get performance gains. Therefore, proposed work follows this line of study, however, with objective of optimizing the execution of database applications in virtualized environments with shared resources, applying a dynamic adjustment policy of disk resources allocation. It aims to distribute disk resources optimally through an algorithm, avoiding that one or more processes consume all disk resources, while others wait to be executed or are being executed without minimum of appropriate disk resources, thus, taking more time to complete their execution. In order to demonstrate this scenario, workloads of OLTP (Online Transaction Processing) and DW (Data Warehouse) databases have been evaluated using the Orion data load simulator [24] and real captured data from a loading test provided by a large IT company in partnership with PUCRS University (Pontifical Catholic University of Rio Grande do Sul), through the Oracle RAT (Real Application Testing) [25]. Laboratory tests have been performed using the following test scenarios: without adjustment of disk resources, with static adjustment of disk resources and through a dynamic adjustment policy of disk resources based on performance metrics. In this case, it can be observed that dynamic policy obtained the best result among the other test groups, generating a gain of 23% for OLTP database workloads, 21% for DW database workloads and 18% for environments with different types of workloads in concurrency like DW and OLTP. |
