سلام
اگر میشه اینو ترجمه کنید
The following four principles drive the design of our data grid architecture. These principles derive
from the fact that data grid applications must frequently operate in wide area, multi-institutional,
heterogeneous environments, in which we cannot typically assume spatial or temporal uniformity of
behavior or policy.
Mechanism neutrality. The data grid architecture is designed to be as independent as possible of
the low-level mechanisms used to store data, store metadata, transfer data, and so forth. This goal
is achieved by de_ning data access, third-party data mover, catalog access, and other interfaces that
encapsulate peculiarities of speci_c storage systems, catalogs, data transfer algorithms, and the like.
Policy neutrality. The data grid architecture is structured so that, as far as possible, design
decisions with signi_cant performance implications are exposed to the user, rather than encapsulated
in \black box" implementations. Thus, while data movement and replica cataloging are provided as
basic operations, replication policies are implemented via higher-level procedures, for which defaults
are provided but that can easily be substituted with application-speci_c code.
Compatibility with Grid infrastructure. We attempt to overcome the diculties of wide area, multi-
institutional operation by exploiting underlying Grid infrastructure [10, 20, 15] (e.g., Globus [9]) that
provides basic services such as authentication, resource management, and information. To this end, we
structure the data grid architecture so that more specialized data grid tools are compatible with lower-
level Grid mechanisms. This approach also simpli_es the implementation of strategies that integrate,
for example, storage and computation.
