RECRIAÇÃO AUTOMÁTICA DE ÍNDICES EM UM SGBD RELACIONAL / AUTOMATIC REINDEXING IN RELATIONAL DATABASES

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

One of the most important tasks of Database Administrators certainly is to guarantee optimal response times to statements submitted by users of big RDBMS. Our dissertation deals with Index Self-tuning, which means creating, dropping or recreating indexes automatically, in order to decrease SQL queries durations. We start from Marcos Salles´ dissertation [32], which proposed an automatic way of creating indexes. We extend [32] in many ways: first using a different workload, TPC-H like. Second, following created indexes inspecting its usage. Finally, we have gotten to drop and, mostly, recreate indexes using different fillfactor in leaf pages. Also, we have elaborated many tests using commercial tools, Microsoft SQL Server 2005 and Oracle 10g in order to ratify [32] ideas. Unfortunatelly, we could not test automatic dropping and recreating in these tools, as long as they do not offer this kind of functionalities. Unlike related work and commercial tools, we have created a code prototype that not only suggests new indexes creations, but also drops and recreates indexes using an own heuristics. To validate our ideas we have used a TPC-C like workload, but we had to make some changes to increase updates and force reindexing.

ASSUNTO(S)

indexes fragmentacao indices self-tuning auto-sintonia fragmentation

ACESSO AO ARTIGO

Documentos Relacionados