Lattice Boltzmann for Colloid Hydrodynamics

The lattice Boltzmann (LB) method provides a very attractiveapproach to the study of mesoscopic systems which have complexgeometry and make-up, and is also well-suited to parallelcomputation. This talk will give an overview of progress madein both these areas as part of the RealityGrid project.

An LB implementation of solid particles (colloids) in a binary fluidmixture has been developed, along with consistent thermal fluctuationswhich are important in mesoscopic systems. Parallel implementationof the solid particles is currently based on shared memory, but thestudy of truly large systems requires a message passing implementation,which is now under development.

Community Grid Services

Grids support (virtual) organizations or communities and we discuss howone can merge Peer-to-Peer and Grid architectures to provide choices in security, scalability and performance. Narada Brokering provides a flexible messaging infrastructure that supports the interactions of clients and Grid Services within a SOAP-basedarchitecture. One can choose protocols and multicast or unicast destinations independently of the payload. GridFTP or P2P messaging are two possible choices. Multicast destinations are used to support collaboration including audio/video conferencing.and we describe the GlobalMMCS system which is a collection of Web services supporting Access Grid, Webcam and Polycom clients. We explain collaborative visualization in this framework.

Experience Building Grid Applications with the CCA Software ComponentModel in an Web Service Framework

The Common Component Architecture (CCA) is a specification for writingsoftware components that can be composed together to build scientific applications. CCA has been used in many large scientific computing applications ranging from computational chemistry to global climate modeling. There are four implementations of CCA and one of those, XCAT,is designed for application that execute in Grid environments. In this framework each component is a Grid Service in that each "port" implements OGSI (and soon WSRF) interfaces. This talk will describe the XCAT model and discuss several of the other component architectures used for building Grid applications. We will also describe the connection of this model of programming to expressing Grid workflow and how Grid portals can be used as a front end to both types of application.
Back

Computational Steering and Real-Time Visualization of MaterialsScience Simulations

This talk will focus on the use and implementation of RealityGrid'scomputational steering tools inconjunction with real-time visualizationfor molecular dynamics (MD) and ab-initio modelling of problems inmaterials science. We will illustrate how these tools can be utilisedfor methodology development, system setup and the discovery of newscience. We will use examples from MD simulations of nanoindentationand nanofriction modelling as well as the simulation of STM imagesusing ab-initio methodologies.
Back

Computational Steering in Monte Carlo Simulations of Thin Film Polymers

High molecular weight polymer systems show very long relaxation times, of the order of one second. This timescale proves practically inaccessible for full scale atomistic dynamical simulation such as molecular dynamics. Even with a Monte Carlo (MC) simulation, the generation of statistically independant configurations is non-trivial.

Many efficient MC moves have been proposed for computer simulation of polymers and each must be parameterised by length scale and selection probabilities chosen for a particular chain length, chemistry and tempaerature. Computational steering provides a new methodology for a systematic search to parameterise trial MC moves for equilibrating a bulk polymer system. We report our progress in the development of a steerable MC code for a freely-jointed bead model and out simulations of thin film polystyrene.
Back

Hybrid particle-continuum modelling of liquids and applications

Complex fluids, such as polymeric or amphiphilic systems, are governed by processes interacting within a broad range of time and length scales, from the nanometre to microns and beyond. Standard simulation techniques are not suited to bridge this gap in scales because they are built to independently describe each complexity level: either the nanoscale, via molecular dynamics (MD), the mesoscale,via dissipative particle dynamics (DPD) or beyond, using continuum fluiddynamics (CFD).

In many cases, the complex region of the system is located near a surface or interface (e.g. membrane) which may exchange mass,momentum, and/or energy with the surrounding fluid. In this talk wepresent a new multiscale technique that can link the several complexity levels within these scenarios. This multiscale method consists on a hybrid scheme that connects a domain described at atomistic level with a region treated via continuum fluid dynamics. The complex interfacial region is solved via MD and connected to a CFD description of the bulk flow, enablingthe exchange of hydrodynamic information from and to the MD region. This hybrid description greatly reduces the computational time, without losing any relevant atomistic information. Our coupling protocol is based on the balance of fluxes of conserved quantities across the particle/continuum interface and it works for unsteady flows involving mass, momentum, and energy transfers.

The particle-continuum hybrid scheme is extremely useful to study complex boundary conditions in micro-fluidics (such as rough surfaces in microchannels), adhesion, wetting, or even crystal growth fromfluid phase (where energy and mass exchanges are dominantprocesses). Most of the applications involve time-dependent flows and to illustrate the applicability of the method we shall present simulations of oscillatory shear flow in microchannels.

We have used the hybrid scheme to study thedynamics of a single tethered polymer under shear flow. The results agree extremely well with standard MD simulationsand with the available experimental data on tethered DNA chains stretchedunder shear flow [Doyle et al. Phys. Rev. Lett. 84, 4769(2000)]. Some comments on this relevant problem and future applications shall be mentioned.

The EGEE Project: The new Grid Infrastructure

In the last few years, the capacity of computing networks has become such as to make the speed of data transfer between local and remote computers more or less the same. At the same time, commodity computing has made even simple personal computers as powerful as computer centres were a few years ago. Furthermore, the web has made people familiar with easy access to distributed computing resources. All these independent and parallel technological developments have led to a vision of a virtual supercomputer centre built by connecting together thousands of computer centres around the world. This could lead to an information technology revolution, the one proposed by the Grid.

The vision of Grid computing is that of an infrastructure which will integrate large, geographically distributed computer clusters and data storage facilities, and provide simple, reliable and round-the-clock access to these resources. For various scientific applications, the benefit will be a large increase in available computing power and peak capacity. Grid technology will also allow scientific communities to develop new ways to share and analyse very large data sets, to the benefit of both the quality and quantity of scientific output.

In recent years, a number of projects have demonstrated first results for various aspects of Grid computing. However, as yet, there are no real production-quality Grids that can offer continuous and reliable Grid services to a range of scientific communities.

The new European project EGEE (Enabling Grids for E-Science in Europe) aims to fulfil this need. Its aim is to integrate current national and regional Grid efforts, and create a seamless European Grid infrastructure for the support of the wider international scientific community. Built on the EU Research Network GEANT, the project will enable round-the-clock access to major computing resources, and will exploit Grid expertise generated by projects such as the EU DataGrid and many other EU, national and international Grid projects.

Integrative Biology: Project Overview

I will give an overview of the aims and objectives of this second round EPSRC Pilot Project. The project aims to build on the achievements of the first round of UK e-Science projects to build a second generation Grid to support the work of experimental scientists. The project will focus on developing the mathematical modelling tools, HPC-enabled simulations and the underpinning Grid infrastructure required for an integrative approach to the modelling of complex biological systems. The two applications on which the project will focus are the modelling of heart disease and cancer.
Back

The DEISA Project and its Joint Research Activities

An overview of the DEISA project, its objectives and the outline of activities are presented. Joint Research Activities in different scientific key areas are described, with a focus on materials sciences and fusion-orientated energy research.
Back

The Software Infrastructure of RealityGrid: Achievements and Prospects

This presentation covers the software infrastructure of RealityGrid. I review our achievements to date, summarise our current capabilities and present the roadmap for software infrastucture developments during the closing stages of the project.
Back

OGSI::Lite - Grid Services in Perl

OGSI::Lite is an implementation of the Open Grid Services Infrastructure in Perl. Grid Services developed in OGSI::Lite can be implemented in a single script as a stand-alone service, they can be hosted in the OGSI::Lite Container, they can be hosted in Apache or they can be implemented using a database to provide persistence: the choice is up to the developer. The priority of OGSI::Lite is to make it as easy as possible to create and customise Grid Services and the Grid Service hosting environment for the developer.
Back

e-Scientists' Real Requirements for Middleware

This talk will describe the current plans for the OMII and will discuss recommendations for what OMII should be doing.
Back

Grid-based Performance Control for Scientific Coupled Models

At the 2003 RealityGrid Workshop, we presented a design for a system capable of steering the performance of an application composed of malleable components executing in a distributed fashion on a computational Grid. This system has been mostly implemented, and the practical question now arises as to how to develop appropriate component-based software that can take advantage of it.

This year we will describe a new scheme, known as the General Coupling Framework (GCF), for generating component-based, performance controllable parallel code from scientific coupled models. The talk will develop a simple scientific coupled model in GCF, and then show how the resulting components may be deployed, and their performance controlled, in a Grid-based execution environment. The talk will include details of the implementation of the associated performance control system (PerCo).
Back

The MUST project: Multicast communication for the Grid

IP Multicast is the enabling technology for efficient point-to-multipoint transfer of information (files, audio and video, resource discovery and search queries) over wide area networks. The MUST project is a collaboration between BT Research and RealityGrid. The aim of the project is to explore the application of multicast to such point-to-multipoint Grid transport applications.
Back

Computer simulations in drug design, and the Grid

Computer simulations are becoming an increasingly important tool in the area of rational drug design. In this presentation, the role of computer simulations in this context will be briefly discussed. The role of Grid computing in this work will also be considered, taking the comb-e-chem and BioSimGrid projects as exemplars. Finally, a new e-science project involving the integration of simulation information across different length and timescales (IntBioSim) will be described.
Back

Grid Computing: Illusion or Reality?

To answer this question, we must first have a clear understanding of what is meant by grid computing. The definition I advocate is:

Grid computing is distributed computing performed transparently across multiple administrative boundaries.

Here, "computing" is to be understood in its widest sense, while the role of transparency is critical to any serious and sustained uptake of grid computing by putative users. Transparency implies ease of use of a persistent grid infrastructure. So far, most grid activities reside in the demonstration class, and are anything but transparent. It is becoming urgent that this situation is ameliorated.

I shall discuss activities within the RealityGrid project which aim at providing a usable grid infrastructure for scientific research. Some of this work is being done in the UK while other aspects are international in character, most notably with US colleagues and the US TeraGrid. I conclude with a critique of current grid middleware and describe a newly funded EPSRC project which is designed to improve the situation for users.
Back

ICENI: An Integrated Grid Middleware to enable e-Science

ICENI (the Imperial College e-Science Networked Infrastructure) provides an end-to-end middleware to support e-Science from the perspective of both the user, the resource providers and the managers of the virtual organisation. ICENI allows an e-scientist to define and compose the workflow within and around their application by supporting the discovery and selection of services from the fabric resources, and the monitoring, visualisation and steering of a running application. Eliminating the barriers provided by different technologies at each stage of this process is key to reducing barriers to entry within the Grid environment.

The latest developments introduced into ICENI to support the RealityGrid project will be reported upon. This includes a persistent testbed featuring run-time visualisation, steering, collaboration and scheduling based around the ‘fast-track’ steering enabled LB3D application which can be accessed through a light-weight Java based client that is easy to install. The usage scenarios within the RealityGrid project demonstrates how a meta-data rich service oriented architecture and augmented component framework that incorporates performance information allows the most appropriate resources to be selected to meet the user’s and resource provider’s specified needs. Further information can be found here.
Back

Simulations of methanol-water mixtures: complex behaviour from simple systems

Methanol is the simplest molecule to contain both hydrophobic and hydrophilic groups. Despite it simplicity, it appears to show prototypical amphiphilic behaviour in aqueous solutions, associating via hydrophobic group contact over a wide composition range. This microimmiscibility -segregation and clustering of both solvent and solute - is now being considered as the origin of the anomalous thermodynamic properties of these solutions, challenging the long-held view, dating back half a century, that this is due to enhanced water structure around hydrophobic groups.

This talk will present results of classical molecular dynamics simulations of different compositions of methanol-water mixtures, characterising the structure and dynamics of the clusters formed and discuss the influence of pressure and temperature on the observed microimmiscibility.
Back

VISIT and UNICORE: Computational Steering meets Grid Computing

Computational steering allows scientists and engineers to gainimmediate visual insight into a running simulation and to modifysimulation parameters accordingly and thus adds a new quality to theefficient usage of HPC-systems. The combination with Grid andcollaborative techniques steering extends these capabilities todistributed teams. VISIT, the 'VISualization Interface Toolkit', is alight-weight library for online visualization and computationalsteering that has been developed at ZAM. It supports bi-directionaltransparent data-exchange between a simulation and one or morevisualization applications simultaneously. Recently, the integrationof VISIT with the UNICORE Grid system has been demonstrated at SCGlobal 03. It adds the ability to VISIT to seamlessly and securelyaccess simulations running anywhere in the Grid and includes basicsupport for collaborative steering. This presentation will give anoverview over VISIT, its current interface to UNICORE and discussplans to develop a Grid Services based version of VISIT. It will closewith an online demonstration of a plasma simulation controlled by VISIT.
Back

Vortex Core Identification in Viscous Hydrodynamics

Vortex cores are important coherent structures of three-dimensional fluid flow. Their robust identification and the elucidation of their dynamics constitute important ongoing areas of hydrodynamic research, and require large-scale computational fluid dynamics simulations. We carry these out using lattice-Boltzmann and pseudospectral codes, supplemented with sophisticated preprocessing and postprocessing methodologies involving Ginsburg-Landau evolution equations of various kinds. We describe how the effective application of these techniques to discover new science necessarily involves the dynamic resizing, remapping, and migration of computational lattices for various purposes. It also involves frequent human intervention via computational steering. We provide an overview of this application with emphasis on the requirements for effective Grid implementation that we believe to be generic to most lattice-based applications. In particular, we describe a remapping library for array subsections and relate how it is used in these studies.
Back