Database Management System Performance Improvement Using Data Mining Techniques

Abstract

In recent times, Cloud computing has grown in popularity due to its lesser cost, scalability newlineand pay-as-you-go model and is expected to grow further in the coming decades newlinewith the advent of technologies like Internet of Things (IOT) in modern societies. Nowa- newlinedays many database and datawarehousing systems are deployed on the cloud and access newlineto these applications is provided as a service (referred to as Database-as-a-Service newline(DBaaS)) which is delivered to the clients from the cloud provider s web servers. newlineThese databases installed on shared environments like cloud, make better use of the newlineresources to ensure improved accessibility, fast automated recovery from failures, automated newlineon-the-go scaling, minimal investment and maintenance and potentially much newlinebetter performance. newlineThe focus of the work presented in this thesis is to improve the performance of databases newlinein shared environment by optimizing the transactional and analytical workload queries newlineusing database cache optimization and view selection techniques. newlineEspecially, in systems like Decision Support Databases or Business Intelligence (BI) newlinethere can be a large number of analytical queries, many of them can be extremely complex, newlineneed to access millions of records, perform multiple scans and joins on numerous newlinelarge sized tables and frequently aggregate and consolidate data. Usually these queries newlinewill have large number of common aggregate subqueries. Moreover, queries with aggregations newlinecannot be fully optimized by using the techniques like indexing. newlineIn a correlated nested query, a subquery contains a reference to a value obtained from newlinea tuple of a higher level query. Such a subquery will be reevaluated for each tuple of newlinethe higher level correlated query. If such a sub-query contains some uncorrelated component, newlinethen we can considerably reduce the query cost by avoiding the re-evaluation newlineof uncorrelated component at each invocation of the subquery by creating its materialized newlineview. In order to optimize such OLAP (Online Analytical Processing) workloads,the first pr