BARROS, A. G.; http://lattes.cnpq.br/0744861260869458; BARROS, Abmar Grangeiro de.
Resumo:
Due to the lack of simulation standards in the context of grid computing, the vast majority
of the researchers end up writing their simulation tools from the scratch, and in an ad-hoc
basis. That leads to an extra effort of design, implementation and validation. Each paper that
presents a new simulator not only carries the risk of validation flaws, possibly avoided with
the reuse of a mature simulation tool, but also delivers a simplistic model for the application
or platform level, which is unlikely to be reused or reproduced in future works. This fact
can be noticed in most of the OurGrid publications, a middleware that enables the creation
of peer-to-peer computational grids, which is developed in the Distributed Systems Lab of
this very University. Many authors write different simulators in order to evaluate different
aspects of this middleware. The objective of this work is to present the implementation techniques
of a discrete simulator to the OurGrid. starting from a simulation model that should
be generic enough to be easily reused in different levels of abstraction and functionality.
That means that the model must comprise most of the aspects of the real system and must be
easily configurable to the different levels of research needs. We validated the proposed simulation
model against completeness, accuracy, scalability, speed, extensibility, and testability
requirements. Finally, we present some use cases that depict the extension of this model for
specific research needs.