Abstract:
Digital preservation aims at maintaining digital objects
accessible over a long period of time, regardless of the
challenges of organizational or technological changes or
failures. In particular, data produced in e-Science domains could
be reliably stored in today’s data grids, taking advantage of the
natural properties of this kind of infrastructure to support
redundancy. However, to achieve reliability we must take into
account failure interdependency. Taking into account the fact
that correlated failures can affect multiple components and
potentially cause complete loss of data, we propose a solution to
evaluate redundancy strategies in the context of heterogeneous
environments such as data grids. This solution is based on a
simulation engine that can be used not only to support the
process of designing the preservation environment and related
policies, but also later on to observe and control the deployed
system.