A TAXONOMICALLY DISTRIBUTED DATA RETRIEVAL MODEL

Paper: 278
Session: E (talk)
Speaker: Ramos Pollan, Raul, CERN, Geneva
Keywords: networking, data interchange, desktop computing, software tools, hierarchical storage management


A TAXONOMICALLY DISTRIBUTED DATA RETRIEVAL MODEL


Miguel Marquina, Raul Ramos Pollan, Arnaud Taddei
CERN, CN Division, DCI Group
1211 Geneve 23, Switzerland


Abstract


As many arising directory services are being implemented and a
variety of collections of data are being made available, it
becomes increasingly necessary for users the need of guidance
through such a network of heterogeneous data. At CERN,
different Physics Collaborations are setting up systems to
provide data related to themselves (people, experiments,
applications configurations, etc.). With this variety of
services, when a generic CERN user wants to find a very
determined piece of data (for instance ``event number 24971
temperature conditions''), what service should he ask to? We
present here a model aimed at solving this Data Retrieval
problem. This is done by proposing a protocol for computers to
exchange information about the data they offer and setting up
the basic mechanisms to build up, maintain and query a
distributed taxonomy of data. This taxonomy will be used to
find a set of locations where a given user's query is likely
to retrieve significant information.

Our current work is being done based on the following points:
1) refining the model, 2) defining an implementation strategy
and 3) studying its feasibility, this is, looking for a
minimal set of conditions (locality, scope, resource
consumption, etc.) under which the model is practical and
functional.