Taxonomia para soluções de balanceamento de carga em sistemas baseados em tabelas Hash distribuídas

• Ano de defesa: 2006
• Autor(a) principal: Andre Ribeiro da Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Tabelas Hash; Taxonomia; Algoritmos; Organização de arquivos (Computação); Processamento paralelo (Computadores); Computação; Hashing (Computação); Processamento eletrônico de dados Processamento distribuído; Redes de computadores
• Link de acesso: http://hdl.handle.net/1843/SLBS-6Y5J6Q

Distributed computational systems are a valuable tool for distributingand processing information and are employed in the solution of severalproblems. A particular problem is the distribution of resources amongpeers of a network. This is not a concern for applications based onembarrassingly parallel algorithms. Many other applications, however,should be able to determine which node has a specific information or could perform a certain task. Depending on the application to be considered, this functionality should be designed to support network churn with minimum alteration in the system. This is the case of the consistent hash technique which corresponds to the main idea behind networks based on Distributed Hash Tables (DHT) that are the base of several structured peer-to-peer networks. An important issue of DHT networks is its inherent load unbalancing due to the way in which peer identifiers are created. This problem has received much attention recently and, as a consequence, several algorithms have been proposed. In order to organize all this knowledge, compare the algorithms proposed in the literature, and provide a base for the development of new strategies, we propose in this work a taxonomy for load balancing algorithms for DHT Networks. As a result of this study, we analyzed several works and classified them according to the proposed taxonomy.