| Monday, October 15 |
| 12:00 pm - 12:30 pm | |
| |
The Web, HPC and UNICORE
(30 mins)
Web 2.0 and Grids
The Web, HPC and UNICORE
R. Menday, B. Hagemeier, A. Streit
r.menday@fz-juelich.de
UNICORE is an established Grid platform providing access to distributed computational resources, particularly HPC resources. Over the past 10 years we have seen UNICORE in production usage at many supercomputer centres and used in many projects. In summer 2007 the final version of the WS-RF compliant, OGSA based UNICORE 6 [1] is released. Building on the core values of UNICORE - "seamless, intuitive and secure access" - in this OGF session we propose to explore the opportunities from closely aligning the Grid and the Web to build a newthe next generation of UNICORE. We intend to give an introduction to the key concepts and a short demonstration of a proof of concept implementation during the session.
This deep integration leads to the embedding of the Grid into the larger Web (more precisely the Semantic Web). As such the Web can be seen as the global, homogenizing middleware layer for the Grid. A RESTful use of HTTP and a unified model for the information of the Grid, result in a loosely-coupled and internet-scale architecture. Each grouping of computational resources (normally aligned with administrative boundaries) have a projection onto the Web. The URI space for a single site is partitioned according to the resources in the site, maintaining a single point of access and the virtualisation of these resources. Viewed from a global perspective, this proposed new UNICORE Grid consists of a distribution of Grid Web sites.
We take a resource-oriented approach where each significant resource (Sites, Users, Files, Jobs, etc) are named with HTTP URIs, and from which various representations can be obtained (RDF, JSON, HTML). Following the recommendations of LinkingOpenData [21] initiative this information about the Grid of computational resources adds to the Web of Data. Resource and data management is performed through the manipulation of this information. Knowledge representation offered by the Semantic Web stack of technologies provides the basis for additional information to be attached to the resources of the Grid. For example, 'tagging' items in the Grid and publishing other additional domain-specific data, and this is a basis for the participatory Grid, offering for example, the selected sharing of jobs and data. Other sites perform an aggregating role (similar to Google search or Yahoo pipes) providing users with a high-level view of their distributed resources and tools to interact with them. For example, searching functionality across this information using the Semantic Web query language SPARQL, providing virtualised views over distributed data, and finally visualisation using timelines and maps.
[1] http://www.unicore.eu
[2] http://esw.w3.org/topic/SweoIG/TaskForces/CommunityProjects/LinkingOpenData
Location: Amphitheater
|
| |
| |
|
| |
| | Link: Slides in HTML |