Speaker
Jan Svec
(CESNET)
Description
New scientific communities, attracted by successful deployment of Grids by
various scientific groups, bring new requirements on Grid middleware, such as
fast turn-around for short jobs, access prioritisation, and need for various
execution environment optimized to application specific requirements.
We will argue that virtualized Grids, based on idea that different virtual
worker nodes, running on and sharing physical cluster nodes, may help to
implement some of these features. In this presentation, we will present
our effort in virtualization of the national grid infrastructure
in Czech Republic and we will describe a system called Magrathea, which we
have developed to allow Grid job scheduling system to deal with several
virtual machines (VMs) running concurrently on a single computer.
With help of two HEP and EGEE related deployment scenarios
(service consolidation of EGEE services and sharing of worker nodes between
EGEE and MetaCentre grids) we will demonstrate how virtual machines can be
already used to fullfill some of new requirements.
| Approximate Length (in minutes) (default is 15 minutes for Site Reports and 30 minutes for others, including questions) | 15 |
|---|
Primary author
Mr
Miroslav Ruda
(CESNET)
Co-authors
Jan Svec
(CESNET)
Dr
Ludek Matyska
(CESNET)