Information Technology Reference
In-Depth Information
the Science and Technology Facilities Council to run a UK grid for particle
physics. Its 17 sites and 5000 computers across the United Kingdom are
part of an even larger international grid that will analyze data from
Europe's next particle accelerator, the Large Hadron Collider [66]. NGS
aims to provide the United Kingdom researchers with computation and
data resources and facilities to help carry out their research, regardless of
actual locations of resources and researchers. It currently pools resources
from nine locations across the United Kingdom. Examples of the UK
e-Science project in molecular simulation include e-Minerals [69] and the
MaterialsGrid [70]. The vision of the e-Minerals project aims to combine
developments in atomistic simulation tools with emerging grid comput-
ing technologies in order to stretch the potential for undertaking simula-
tion studies under increasingly realistic conditions, and which can scan
across a wide range of physical and chemical parameters [69]. The aim of
the MaterialsGrid project is to create a pilot dynamic database of materials
properties (such as elastic stiffness, dielectric constants, optical properties,
heat capacity, and electronic band gap) based on quantum mechanical
simulations run within Grid computing environments [70]. More about
MaterialsGrid will be discussed in the next two sections.
11.4.2
MaterialsGrid: Large-Scale Simulation of Physical
Properties of Materials
MaterialsGrid is a UK government DTI-funded e-Science research project
($5.4 million) that aims to create a unique database of critical electronic
and physical materials properties based on highly accurate quantum
mechanical simulation methods using the power of grid computing.
Scientists can search the database, retrieve precomputed properties where
available, or trigger new calculations that i ll in missing data. Users will
be able to donate data and grid resources in exchange for access, ensuring
the continued growth of the service.
Running a quantum mechanical simulation for material properties
(e.g., CASTEP, SIESTA) on the grid typically involves the following
steps: (1) creating the simulation input i les, (2) copying the input i les
and simulation code to remote computational resources, (3) logging
into the computational resource and submitting the simulation job, (4)
waiting for the job to i nish, and once the job i nishes copying back the
simulation output i les to the local machines, and (5) harvesting the
data or metadata and storing them to the material property database.
This approach works successfully but it does have disadvantages: i rst,
the approach involves a lot of human interaction; second, in order
to submit job(s) to remote resources, some grid software has to be
installed (e.g., Globus) in a local machine, or logging into a machine
where such grid software has been installed is necessary; third, it will be
Search WWH ::
Custom Search