HR: 1330h
AN: U22A-0010    [PDF]
TI: Virtual Observatories: A Peer-to-Peer Semantic Approach for Efficient Sharing of Scientific Data
AU: * Saxena, A
EM: absaxena@umich.edu
AF: AOSS/SPRL, The University of Michigan, Ann Arbor, MI 48109 United States
AU: Papitashvili, V
EM: papita@umich.edu
AF: AOSS/SPRL, The University of Michigan, Ann Arbor, MI 48109 United States
AB: A recent initative on the {\it electronic Geophysical Year (eGY)} calls for the establishment of a series of {\it virtual geophysical observatories} now being ``deployed'' in cyberspace. The {\it eGY} concept creates the means for a free and transparent data access to remote, worldwide distributed databases through the Internet and World Wide Web. However, current forms of sharing scientific data are either highly centralized or require intensive personal communication. Centralized distribution schemes need continuing maintenance and support that somewhat counteracts the rationale of a free access to such data repositories. Here we argue for the development of a {\it peer-to-peer} semantic approach focused towards efficient sharing of scientific data via the Internet. Our proposed data distribution model aims to simplify the process of data acquisition and storage by tapping the vast potential of end user machines. At the same time, the data integrity and publisher accountability should be an integral design goal of any platform that aims to provide distributed, yet reliable, means of sharing such high-implication data amongst a rapidly increasing and widespread scientific community. Current design philosophies of the file-sharing networks model the data files as ``once injected, never modified'' objects, which are shared through a set of pre-specified interfaces. On the contrary, scientific data sets are often multidimensional and the same object can be published through multiple dynamically changing interfaces. Furthermore, from time to time the user-specific analyses of these data often generate additional summaries and even new datasets. Publishing these derived datasets may introduce more complexity and challenges, but the appropriate authentication and signing of the ``original'' and ``modified'' datasets can resolve the problem. We explore the possibility of the semantic-oriented information modeling of the scientific data objects for enhancing real-time, machine-to-machine search capabilities of the proposed scheme.
UR: http://maggy.engin.umich.edu/mist/vgmo.html
DE: 0399 General or miscellaneous
DE: 1699 General or miscellaneous
DE: 1899 General or miscellaneous
DE: 2799 General or miscellaneous
SC: U
MN: 2003 Fall Meeting