RepliX: Um mecanismo para a replicaÃÃo de dados XML

• Autor(a) principal: FlÃvio Rubens de Carvalho Sousa
• Data de Publicação: 2007
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFC
• Texto Completo: http://www.teses.ufc.br/tde_busca/arquivo.php?codArquivo=1361
• topic: BDs XML Nativos; comunicaÃÃo em grupo, replicaÃÃo; Native XML Databases; group communication; replication; CIENCIA DA COMPUTACAO

XML has become a widely used standard for data representation and exchange in applications. The growing usage of XML creates a need for ecient storage and recovery systems for XML data. Native XML DBs (NXDBs) are being developed to target this demand. NXDBs implement many characteristics that are common to traditional DBs, such as storage, indexing, query processing, transactions and replication. Most existing solutions solve the replication issue through traditional techniques. However, the exibility of XML data imposes new challenges, so new replication techniques ought to be developed. To improve the performance and availability of NXDBs, this thesis proposes RepliX, a mechanism for XML data replication that takes into account the main characteristics of this data type, making it possible to reduce the response time in query processing and improving the fault-tolerance property of such systems. Although there are several replication protocols, using the group communication abstraction for communication and fault detection has proven to be a good solution, since this abstraction provides ecient message exchanging techniques and conability guarantees. RepliX uses this strategy, organizing the sites into an update group and a read-only group in such a way that allows for the use of load balancing among the sites, and makes the system less susceptible to faults, since there is no single point of failure in each group. In order to evaluate RepliX, a new replication layer was implemented on top of an existing NXDB to introduce the characteristics of the proposed mechanism. Several experiments using this layer were conducted, and their results conrm the mechanism's eciency considering the dierent aspects of a replicated database, improving its performance considerably, as well as its availability.