Castiglioni, A. (2011):
Assessment of dependability scenarios with failures in large scale grids using GridSim toolkit
Grid networks are used to process large amounts of data and use a large number of resources. These resources are embedded in virtual organizations and in coordination with each other, share the computing power to execute certain jobs. Grid systems must exhibit a high degree of dependability, i.e., the ability to appear trustworthy to users. A essential feature of grid networks is their large scalability, which represents a major challenge to understand the "dependability". In this project, we simulate large scale grids using the GridSim toolkit provided by the University of Melbourne. This toolkit is modified to adapt to our needs of having a virtual grid infrastructure on a large scale, in which the components (resources, routers, machines, etc.) can have failures of any kind: natural disaster, malicious attack or normal faults. We study with controllable, repeatable and observable experiments the behavior of the grid network when a single point of failure occurs and when a set of failures affects the grid network. The results of these simulations are compared with the results of the same architecture grid network without failure, to evaluate the differences of behaviour, and to understand the dependability of the network.