InteGrade: Um Sistema de Middleware para Computação em Grade Oportunista / InteGrade: A Middleware System for Opportunistic Grid Computing

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

The past years witnessed a substantial increase in the need for computing power in various fields of human activity, including many industrial and academic endeavors. Grid Computing addresses those needs, providing seamless access to distributed computing resources, allowing one to use the combined computing power of various machines. However, the majority of the earlier Grid Computing systems focused on connecting high performance computers, which are very expensive resources only accessible to a small number of institutions. Contrasting with high cost parallel computing, personal computing experienced a tremendous growth in the last fifteen years. Personal computers are ubiquitous, cheap, and extremely powerful. The increase in processing power motivated the creation of dedicated PC clusters, allowing one to perform high performance computing tasks at a fraction of the price of a traditional parallel machine. Although cheaper, building a cluster still requires a considerable investment. At the same time, institutions that rely on the processing power of dedicated clusters typically own a large number of personal computers that are idle for most of the time, resulting in a loss of computing power that could otherwise be used for computing tasks. Opportunistic Grid Computing systems allow the use of the idle computing power of personal computers to perform useful computation. Many Opportunistic systems were successfully employed to solve problems in areas such as astronomy, biology, and mathematics. InteGrade, an Opportunistic Grid Computing system developed in the context of this thesis, aims to provide features not commonly available in other Opportunistic systems, such as support for parallel applications that require communication among application nodes, and usage pattern collection and analysis, which will allow for better scheduling decisions by providing predictions about future resource availability. InteGrade is a fully object oriented system, featuring both object oriented architecture and implementation. The work presented in this thesis includes a survey of existing Grid Computing systems and the definition of the InteGrade initial architecture, including the specification and implementation of various software modules. We also present the design and implementation of a parallel programming library that implements the BSP computing model, which allows one to write parallel applications that execute on InteGrade.

ASSUNTO(S)

distributed systems sistemas distribuídos grid computing corba corba computação em grade

ACESSO AO ARTIGO

Documentos Relacionados