Speaker
Mr
Riccardo Zappi
(INFN-CNAF)
Description
In the near future, data on the order of hundred of Petabytes will be spread in
multiple storage systems worldwide dispersed in, potentially, billions of replicated
data items. Users, typically, are agnostic about the location of their data and they
want to get access by either specifying logical names or using some lookup mechanism.
A global namespace is a logical layer that allows the view of data resources
independently from the physical location. Usually, the naming scheme is designed to
be easily interpreted by humans and it is organized into a purely user-defined
directory hierarchy. Within this model, a data resource is uniquely addressed by file
name and path. Nevertheless, this hierarchical structures of logical namespace lacks
adequate flexibility to manage sophisticated organization of data. In particular the
implicit classification of the data item derived from the path is not enough
meaningful to classify data objects when different orthogonal dimensions are considered.
In this paper we expose an enriched namespace able to support a new type of data
access authorization policy based on tags. The tags are organized in well-defined
hierarchies providing a simple representation of the domain ontology. Only authorized
users can label data resources with different tags taken from the domain tag
hierarchies. In this way an overlay of classical hierarchical structure of logical
namespace with faceted hierarchical tags provides a semantics classification of data
entities. Authorization policies defined in respect of tags are content-aware.
Authors
Dr
Alberto Forti
(INFN-CNAF)
Prof.
Antonia Ghiselli
(INFN-CNAF)
Mr
Federico Stagni
(INFN)
Mr
Luca Magnoni
(INFN-CNAF)
Mr
Riccardo Zappi
(INFN-CNAF)
