Information about http://www.eminerals.org/papers/MolSim_vision_2005.pdf
Molecular Simulation, Vol. 31, No. 5, April 2005, 297301 …
Tags: birkbeck college, cambridge cb2 3eq, cambridge cb3, computing resources, crystallography, de leeuw, earth sciences, grid infrastructure, london wc1e 7hx, malet street, molecular simulation, molecular simulations, monte carlo simulations, performance resources, simulation studies, street cambridge, street london, university of cambridge, virtual organization, wilberforce road,Pages: 5
Language: english
Created: Mon May 30 09:07:16 2005
Display cached document
Page 1
Page 2
Page 3
Page 4
Page 5
Molecular Simulation, Vol. 31, No. 5, April 2005, 297301
Grid computing and molecular simulations: the vision
of the eMinerals project
M. T. DOVE.,* and N.H. DE LEEUW§*
Department of Earth Sciences University of Cambridge Downing Street Cambridge CB2 3EQ UK
National Institute for Environmental eScience, Centre for Mathematical Sciences University of Cambridge Wilberforce Road Cambridge CB3 0EW UK
§School of Crystallography, Birkbeck College Malet Street London WC1E 7HX UK
(Received November 2004; in final form November 2004)
This paper discusses a number of aspects of using grid computing methods in support of molecular simulations, with
examples drawn from the eMinerals project. A number of components for a useful grid infrastructure are discussed, including
the integration of compute and data grids, automatic metadata capture from simulation studies, interoperability of data
between simulation codes, management of data and data accessibility, management of jobs and workflow, and tools to support
collaboration. Use of a grid infrastructure also brings certain challenges, which are discussed. These include making use of
boundless computing resources, the necessary changes, and the need to be able to manage experimentation.
Keywords: Grid computing; eScience; Virtual organization; Monte Carlo simulations
1. Introduction the early efforts were based on linking together high-
performance resources, but the biggest impact is coming
Molecular simulation scientists have always needed from producing larger grids of more modest resources. One
significant computing resources. In the early days, example is the approach of linking together hundreds, or
limitations on available computing power led to a number even thousands, of under-used desktop machines. Consider
of constraints, such as the size of the system that could be the example of forming a grid across a campus, and linking
studied, the length of time for which a simulation could be together 1000 under-used PCs. If each PC operates at
run, the complexity of the forces that could be used, and around 5 Gflops (a quick test shows that the author's laptop
the number of state points (e.g. temperature) that could be is running at 4.3 Gflops) and has 1 GB memory, a quick
used within a single study. Over the years, computer power multiplication shows that the grid resource has 5 Tflops of
available to computational scientists has grown at an power with 1 TB memory. This would equate with
exponential rate, both at the desktop and at high- computers around the 30th place in the world rankings
performance computing facilities, greatly increasing the (www.top500.org). Of course, a grid of computers does not
horizons of the simulation scientists. Growth in computing make a high-performance computer because connectivity
capacity has been matched by developments of new between separate nodes will not be fast enough to support
simulation methods and algorithms. Similarly, desktop parallel applications, there can be no guarantees against
visualization tools have become widely available to enable loss-of-service for individual nodes, and the machines on
the simulation scientist to view the results of a simulation the grid are likely not to be of homogeneous hardware and
through graphical manipulation tools and through with common operating systems. Less technical, it is also
animations. likely that different machines within a computing grid will
The development within the past decade of grid be subject to different usage policies. Such a system is more
computing and e-science methods [1 4] offers the prospect appropriate for Monte Carlo simulations, for example,
of new evolutionary (even revolutionary) developments in where each node will gather a subset of the total
molecular simulations. Grid computing was developed configurations needed in an individual study, or for high-
with the idea of linking together computing resources to throughput studies where each node performs an
create large-scale computing infrastructures. Some of independent calculation. An example of the latter could
*Corresponding author. E-mail: martin@esc.cam.ac.uk
Molecular Simulation
ISSN 0892-7022 print/ISSN 1029-0435 online q 2005 Taylor & Francis Group Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/08927020500065801
298 M. T. Dove and N. H. de Leeuw
be a simulation study over many temperatures. An example molecular level with increased levels of realism, achieved
of the use of a grid environment to accumulate statistics and through being able to use larger samples, more complex
temperatures is a study of diffusion of ions through systems, or through increased accuracy to better handle
interfaces by Monte Carlo methods [5]. changes in chemical bonding [26]. The science appli-
Grid computing gives the prospect of what has been cations concern issues such as nuclear waste encapsulation
described as "boundless computing" [6] in that there is a [16,17,27,28], adsorption of organic pollutants on soil
considerable wealth of underused computing power in particles [18,19], and surface processes, weathering and
many institutes [7,8]. The default configuration for many precipitation of minerals [29 33]. The wider aim of the
PCs is equivalent to the supercomputing capabilities of project is to develop a cross-institute collaborative
just a few years ago in terms of processing speed and infrastructure that can be used by the scientists to
processor memory, and hence these machines are capable accomplish the scientific objectives [20].
of being used as compute engines for molecular Based on the discussion in the introduction, the
simulation studies (discussed above). The challenge for eMinerals project team has needed to take a three-track
the molecular simulation scientist is to be able to harness approach, based on the science to be accomplished
this computer power. (including the development of the simulation codes), the
The original vision of grid computing [1] was grid infrastructure that has been needed to support the
characterised by the need to link together computing science, and the virtual organization tools required to
resources, but there is a much wider vision. The challenge facilitate the team working together. The project team
of managing simulations performed over a grid resource consists of a mixture of scientists, code developers,
has to be met through the use of job submission, computer scientists and grid specialists. To some extent the
monitoring and workflow tools. With increased computing project has been an experiment in enabling such a diverse
resources is also the need for data management methods. team to work together towards a common objective.
Running a large study with grid resources will necessarily The simulation work in the eMinerals project involves a
lead to the generation of many output files, too many to be wide range of methodologies. These include both models
able to manage using conventional approaches. The data with classical empirical potentials and quantum mech-
management challenge is further compounded by the fact anics, the latter employing both plane wave and localized
that simulations performed on a distributed compute basis sets. The quantum mechanical methods include both
facility will lead to data being stored in several locations. density functional and Quantum Monte Carlo approaches.
With the emergence of the concept of grid computing Both methods are used for static energy and lattice
arose the concept of virtual organisations (VOs). This dynamics methods (zero-temperature methods) and
concept arose as an independent idea, in fact without a molecular dynamics methods (for non-zero temperatures).
necessary dependence on IT, but it is a particularly With respect to the molecular dynamics method, one of
pertinent idea for escience where organisations come our challenges has been to develop a molecular dynamics
together to share computing resources [9,10]. In fact, code that can handle the necessarily large samples
collaborative working is increasingly becoming more of required for radiation damage studies [17,27,28] in
the norm in terms of funding. In order to maximise the systems with long-range electrostatic interactions. This
potential of simulation scientists working within a VO, it has led to the development of the DL_POLY3 molecular
is essential for them to have access to collaborative tools, dynamics code as part of the work of the eMinerals project
such as desktop videoconference and application-sharing [34]. One interesting aspect of the work of the eMinerals
tools. Furthermore, the access to data needs to support project has been to enable code developers to work with
easy data sharing, and it is essential for data to be stored in the end users (i.e. the scientists) and to take advantage of a
a format that can be shared between different applications. heterogeneous computing environment within which to
This paper serves as an introduction to a collection of test the codes.
papers [10 20] written by members of the eMinerals In addition to enabling scientists, code developers and
project [21], which represent an attempt to develop an grid specialists to work together in ways that have not
integrated compute and data grid infrastructure happened before, the eMinerals team also recognizes the
[19,20,22 25] together with support to enable the project capacity within many team members to multitask outside
team to work as a VO [9,10]. The rest of this paper gives a their notional area of expertise. Thus, for example, some of
broad outline of the vision that drives the project. the scientists have become members of the grid develop-
ment team, other scientists have joined with grid specialists
to take a leadership role in developing XML applications,
2. The eMinerals project
The eMinerals project (formal name, "Environment from 3. The vision of the eMinerals project
the Molecular Level") is one of the escience testbed
projects funded by NERC [21]. The scientific goal of the The broad vision is to develop an infrastructure that allows
project is primarily to use grid-computing methods to scientists to do the science they want to do, free of
facilitate simulations of environmental processes at a the limitations of managing resources and data, free
Grid computing 299
of the need to learn new tricks, free of the need to convert analysed, and kept with the data. The approach within
data between different formats, free to collaborate, free to the eMinerals project is to build this in to the portal
share data, and free to share resources. This vision can be infrastructure currently being developed [37].
broken down into several components. 4. Interoperability. One constant problem is that
programs generate data in formats that prevent easy
1. Compute grid infrastructure. Shared computing re-use of the data in other programs. Although there
resources, including the use of resources that may are many instances where data can simply be parsed
otherwise be barely used, can be combined to produce from one format to another, it is also possible that the
an infrastructure of some considerable computing formats of files contain exceptions (such as when
power [1 4]. In fact, given the large amount on programs throw out a helpful line of information when
untapped desktop resources that can be found in a a certain condition has been encountered). Thus
typical institute, there is the prospect of collaborations central to the vision is the need to be able to handle
having access to what has been described as data in a general form that enables the data to be
"boundless computing resources" [6,8]. As noted understood by other programs. Implicit in this is the
below, the possibility of providing boundless resources need for data to be self-described, which links to the
to simulation scientists will provide new challenges. use of metadata. For many of our applications we are
2. Integrated data grid infrastructure. Computing and finding that XML provides the functionality and
data go hand in hand, and the vision for the eMinerals interoperability we require. XML files can readily be
project has at its core the integration of computing converted into a form that can be view via a standard
resources and data management [19,20,22]. When jobs web browser, including assembling data into graphical
run on distributed computing resources, the data form. The vision for the eMinerals project encom-
generated will need to be managed in ways that hide passes support for the simulation scientists to obtain an
the distribution from the user. Moreover, partners immediate view of the results of simulation studies,
within any collaboration will need to be able to access including studies in which many separate jobs are run,
data without needing to be told specific details of the without the tedious process of pasting numbers into
location of files [24,25,35]. The vision for the separate data analysis programs [18,19,38,39].
eMinerals project includes the need to provide an 5. Management of data and data accessibility. This point
infrastructure to support collaborative and grid data follows from the previous points. Collaborators need to
management, and also, as discussed in the following be able to share data easily. Appropriate use of metadata
points, enable data to be understood by partners and by and data mark-up will make the experience of handling
the codes run by partners. shared data easier, but on top of this is the need for a data
3. Automatic metadata capture. Files of data require archiving infrastructure that enables rapid access to the
information about the data [36], such as when the data data [24,25,35]. The same metadata should be used to
were obtained, from which simulation program, the enable collaborating scientists to search through the
person who ran the simulation, and the computer that archives in order to obtain exactly the sets of data (which
the simulations were run on. The file system will may involve many files) that are required [19].
provide a date and file owner, but that information may 6. Management of jobs, including workflow. Simulation
be corrupted through moving the data file between file scientists tend to manage their tasks manually or through
systems. Ideally simulation programs will provide bespoke scripts. A typical study will involve a number of
headers within the data files to provide information discrete stages, including setting up the files, running the
about the program (such as version number), but jobs, monitoring their progress and intervening if
frequently these headers will not be propagated into all necessary, gathering together the output data, extracting
output files. Some of this information may be stored in the key numbers, performing subsequent analysis and
written documentation by the scientist, but ideally data presentation tasks, and archiving the data generated
such information should be stored with the data, not in in the study. This frequently involves the scientist in
a physically different medium [24,25]. One type of carrying out a number of tedious tasks, and the
metadata that is not built into many systems concerns management of complex workflow patterns may often
the context of the data, covering issues such as the be the bottleneck in the scientist's productivity. Many of
motivation for collecting the data (for example, is the the difficulties in handling the job management and
file a result of a production or test run, and is it part of a workflow tasks can be solved using escience methods.
larger series of data files, and in any case why did the 7. Collaborative infrastructure. e-Science is beginning to
scientist want to run this particular simulation?), and provide the tools to enable collaborators based in
the quality of the data (is this the best simulation that geographically distributed locations to be able to work
was possible at the time, or do the results look together as if sharing adjacent offices within a single
suspect?). Again, some of this information may be building. Among the tools being developed are desktop
captured in a notebook, but the vision is for such audiovisual conferencing tools and tools for sharing
metadata to be captured automatically, whether at the applications. Ideally such tools should not have
time of the submission of the job or when the data are significant overheads for the user, and should require
300 M. T. Dove and N. H. de Leeuw
no more effort than that is required to knock on the door areas in which the vision can be developed, but the
of the person in the next office along the corridor. The challenges have yet to be explored. To mention two,
vision is to be able to reproduce the experience of testing the scalability for larger virtual organisations, and
geographical proximity for distributed collaborating inclusion of experimental data. At the time of writing, the
scientists [9,10]. This goes hand in hand with the tools vision of the eMinerals project is still being implemented,
for sharing resources and data. and much progress is still awaited. The collection of
papers following this introduction provides a snapshot of
some of the technological developments that have been
4. The challenge of the vision
made and implemented, and gives an impression of the
science that is being carried out by the eMinerals project
The main problem with the vision outlined above is that it will
team.
require significant changes to the way that scientists carry out
their work. In particular we identify several challenges that
face individual scientists as well as the project team. Acknowledgements
1. Challenge of access to boundless computing. At face We are grateful to NERC for financial support of the
value, this challenge is counterintuitive, since boundless eMinerals projects. We are appreciative of the support of
computing is every simulation scientist's dream. the eMinerals project team.
However, most scientists' working practices mitigate
against making use of boundless computing. Scientists
often prefer to handle one study at a time, and to develop
References
the study incrementally through running simulations in
sequential order. Access to boundless computing enables [1] I. Foster, C. Kesselman (Eds), The Grid: Blueprint for a new
scientists to launch many more simulations that they computing infrastructure, 1st edition, Morgan-Kaufman (1998).
could normally manage, which requires several [2] I. Foster, C. Kesselman (Eds), The Grid: Blueprint for a new
computing infrastructure, 2 edition, Morgan-Kaufman (2003).
components of the vision outlined in the previous [3] I. Foster, C. Kesselman, S. Tuecke. The Anatomy of the grid:
section. What access to boundless computing gives the enabling scalable virtual organizations. Int. J. Supercomput. Appl.,
simulation scientist is the ability to run several projects in 15, 200 222 (2001).
[4] F. Berman, A.J.G. Hey,, G. Fox (Eds). Grid Computing: Making the
parallel, to run many speculative simulations, to obtain global infrastructure a reality, John Wiley, Australia (2003).
many more data points than previously possible, and to [5] M. Calleja, M.T. Dove. Calculating activation energies in diffusion
gain many more data for statistical analysis. The tools processes using a Monte Carlo approach in a grid environment.
Lect. Notes Comput. Sci., 3039, part 4, 483 490 (2004).
enable the simulation scientist to think ahead and launch [6] M. Linvy, From a talk at the UK Condor Week 2004 held at the
jobs for a study to which he/she would like to devote time National eScience Centre in Edinburgh, (2004).
only at some point in the future. [7] D. Thain, T. Tannenbaum, M. Livny. Condor and the grid. In Grid
Computing: Making the global infrastructure a reality, F. Berman,
2. Challenge of change. It is a fact of life that most of us get A.J.G. Hey, G. Fox (Eds), Chapter 11, John Wiley, Australia (2003).
used to a certain way of doing things, and find it hard to [8] P. Wilson, W. Emmerich, J. Brodholt. Leveraging HTC for UK
make the time required to change to a new way of e-Science with very large Condor pools: Demand for transforming
untapped power into results. Proceedings of the UK e-Science All
working. The vision of escience for simulation science is Hands Meeting 2004, pp. 308 315 (2004), (ISBN 1-904425-21-6).
a significant change to the normal way that most [9] M.T. Dove, M. Calleja, R. Bruin, J. Wakelin, M. Keegan, S. Ballard,
simulation scientists will operate, and there is some G. Lewis, S.M. Hasan, V. Alexandrov, R.P. Tyer, I. Todorov, P.
Wilson, M. Alfredsson, G.D. Price, C. Chapman, W. Emmerich, S.
effort required to learn the new ways of doing things. It is Wells, A. Marmier, S. Parker, Z. Du. Collaborative tools in support
not just a case of learning new techniques; it must not be of the eMinerals Virtual Organisation. Proceedings of the UK
underestimated that the challenge is to the very way in e-Science All Hands Meeting 2004, pp. 127 134 (2004), (ISBN
1-904425-21-6).
which simulation scientists normally operate. [10] M.T. Dove, M. Calleja, R. Bruin, J. Wakelin, M.G. Tucker, G.
3. Challenge of experimentation. Grid computing is a long Lewis, S.M. Hasan, V.N. Alexandrov, M. Keegan, S. Ballard, R.P.
way from maturity. Many of the components required to Tyer, I. Todorov, P.B. Wilson, M. Alfredsson, G.D. Price, C.
Chapman, W. Emmerich, S.A. Wells, A. Marmier, S.C. Parker, Z.
properly implement the vision are simply not yet Du. The eMinerals collaboratory: tools and experience. Mol. Simul.,
available in a robust form. Thus scientists engaged in a pp. 329 338, 12345678 (2005).
grid computing environment at the present time need to `
[11] M. Alfredsson, J.P. Brodholt, P.B. Wilson, G.D. Price, F. Cora, M.
Calleja, R. Bruin, L.J. Blanshard, R.P. Tyer. Structural and magnetic
understand that they will most likely suffer some level of phase transitions in simple oxides using hybrid functionals. Mol.
inconvenience through tools not quite doing what they Simul., pp. 367 378, 12345678 (2005).
are required to do! [12] Z. Du, N.H. de Leeuw, R. Grau-Crespo, P.B. Wilson, J.P. Brodholt,
M. Calleja, M.T. Dove. A computational study of the effect of LiZK
solid solutions on the structures and stabilities of layered silicate
materials--an application of the use of Condor pools in molecular
5. Summary simulation. Mol. Simul., pp. 339348, 12345678 (2005).
´
[13] M.V. Fernandez-Serra, G. Ferlat, E. Artacho. Two exchange-
The vision we have outlined above and in this collection of correlation functionals compared for first-principles liquid water.
Mol. Simul., pp. 361366, 1234 5678 (2005).
papers has been developed specifically for a collaboration [14] S. Wells, D. Alfe, L.J. Blanchard, J.P. Brodholt, M. Calleja, C.R.A.
between molecular simulation scientists. There are several Catlow, G.D. Price, R. Tyer, K. Wright. Ab initio simulations of
Grid computing 301
magnetic iron sulphides. Mol. Simul., pp. 379 384, 12345678 K. Trachenko, J. Wakelin, K. Wright. Science outcomes from the
(2005). use of Grid tools in the eMinerals project. Proceedings of the
[15] A. Marmier, H. Spohr, D.J. Cooke, S. Kerisit, J.P. Brodholt, P.B. UK e-Science All Hands Meeting 2004, pp. 240 247 (2004),
Wilson, S.C. Parker. Self diffusion of argon in flexible, single wall, (ISBN 1-904425-21-6).
carbon nanotubes. Mol. Simul., pp. 385 388, 12345678 (2005). [27] K. Trachenko, M.T. Dove, M. Pruneda, E. Artacho, E. Salje, T.
[16] J.M. Pruneda, L. Le Polles, I. Farnan, K. Trachenko, M.T. Dove, E. Geisler, I. Todorov, B. Smith. Radiation-induced structural changes,
Artacho. Calculation of the effect of intrinsic point defects and percolation effects and resistance to amorphization by radiation
volume swelling in the nuclear magnetic resonance spectra of damage. In Radiation Effects and Ion-Beam Processing of
ZrSiO4. Mol. Simul., pp. 349354, 1234 5678 (2005). Materials, L.M. Wang, R. Fromknecht, L.L. Snead, D.F. Downey,
[17] K. Trachenko, M.T. Dove, E.K.H. Salje, I. Todorov, W. Smith, M. H. Takahashi (Eds), Materials Research Society Symposium
Pruneda, E. Artacho. Radiation damage in the bulk and at the Proceedings792, pp. 509 524 (2004).
surface. Mol. Simul., pp. 355360, 12345678 (2005). [28] K. Trachenko, M.T. Dove, T. Geisler, I. Todorov, B. Smith.
[18] J. Wakelin, P. Murray-Rust, S. Tyrrell, Y. Zhang, H.S. Rzepa, A. Radiation damage effects and percolation theory. J. Phys. Condens.
´
Garcia. CML tools and information flow in atomic scale Matter, 16, S2623 S2627 (2004).
simulations. Mol. Simul., pp. 303 310, 12345678 (2005). [29] N.H. de Leeuw, Z. Du, J. Li, S. Yip, j.T. Zhu. Computer modeling
[19] C. Chapman, J. Wakelin, E. Artacho, M.T. Dove, M. Calleja, R. study of the effect of hydration on the stability of a silica nanotube.
Bruin, W. Emmerich. Workflow issues in atomistic simulations. Nanoletters, 3, 1347 1352 (2003).
Mol. Simul., pp. 311316, 12345678 (2005). [30] Z. Du, N.H. de Leeuw. A combined density functional theory and
[20] M. Calleja, R. Bruin, M.G. Tucker, M.T. Dove, R.P. Tyer, L.J. interatomic potential-based simulation study of the hydration of
Blanshard, van Dam K. Kleese, R.J. Allan, C. Chapman, W. nano-particulate silicate surfaces. Surf. Sci., 554, 193 210 (2004).
Emmerich, P.B. Wilson, J.P. Brodholt, A. Thandavan, V.N. [31] A. Marmier, S.C. Parker. Ab initio morphology and surface
Alexandrov. Collaborative grid infrastructure for molecular thermodynamics of a-Al2O3. Phys. Rev. B, 69, 115409 (2004).
simulation: The eMinerals minigrid as a prototype integrated [32] S.C. Parker, S. Kerisit, A. Marmier, S. Grigoleit, G.W. Watson.
compute and data grid. Mol. Simul., pp. 317328, 12345678 Modelling inorganic solids and their interfaces: A combined
(2005).
approach of atomistic and electronic structure simulation
[21] M.T. Dove, M. Calleja, J. Wakelin, K. Trachenko, G. Ferlat, P.
techniques. Faraday Discuss., 124, 155 170 (2003).
Murray-Rust, N.H. de Leeuw, Z. Du, G.D. Price, P.B. Wilson, J.P.
[33] S.C. Parker, D.J. Cooke, S. Kerisit, A. Marmier, S.L. Taylor, S.N.
Brodholt, M. Alfredsson, A. Marmier, R.P. Tyer, L.J. Blanshard,
Taylor. From HADES to PARADISE--atomistic simulation of
R.J. Allan, K. Kleese van Dam, I.T. Todorov, W. Smith, V.N
defects in minerals. J. Phys. Condens. Matter, 16, S2735S2749
Alexandrov, G.J. Lewis, A. Thandavan, S.M. Hasan. Environment
(2004).
from the molecular level: an e-science testbed project. Proceedings
of the UK e-Science All Hands Meeting 2003, pp. 302305 (2003), [34] I.T. Todorov, W. Smith. DL_POLY 3: the CCP5 national UK code
(EPSRC, ISBN 1-904425-11-9). for molecular-dynamics simulations. Philos. Trans. R. Soc. London
[22] M. Calleja, L. Blanshard, R. Bruin, C. Chapman, A. Thandavan, R. A, 362, 18351852 (2004).
Tyer, P. Wilson, V. Alexandrov, R.J. Allen, J. Brodholt, M.T. Dove, [35] G. Drinkwater, K. Kleese, S. Sufi, L. Blanshard, A. Manandhar, R.
W. Emmerich, van Dam K. Kleese. Grid tool integration within the Tyer, K. O'Neill, M. Doherty, M. Williams, A. Woolf. The CCLRC
eMinerals project. Proceedings of the UK e-Science All Hands data portal. Proceedings of the UK e-Science All Hands Meeting
Meeting 2004, pp. 812 817 (2004), (ISBN 1-904425-21-6). 2004, pp. 540547 (2004), (ISBN 1-904425-21-6).
[23] R.P. Tyer, L.J. Blanshard, K. Kleese van Dam, R.J. Allan, A.J. [36] S. Sufi, B. Mathews, van Dam K. Kleese. An interdisciplinary
Richards, M.T. Dove. Building the eMinerals Minigrid. Proceed- model for the representation of scientific studies and associated data
ings of UK e-Science All Hands Meeting 2003, pp. 483487 (2003), holdings. Proceedings of UK e-Science All Hands Meeting 2003,
(EPSRC, ISBN 1-904425-11-9). pp. 103 110 (2003), (EPSRC, ISBN 1-904425-11-9).
[24] L.J. Blanshard, K. Kleese, M.T. Dove. Environment from the [37] R. Tyer, M. Calleja, R. Bruin, C. Chapman, M.T. Dove, R.J. Allan.
molecular level e-science project and its use of CLRC's web Portal framework for computation within the eMinerals project.
services based data portal. Proceedings of the First International Proceedings of the UK e-Science All Hands Meeting 2004,
Conference on Web Services, pp. 164167 (2003), (CSREA Press, pp. 660 665 (2004), (ISBN 1-904425-21-6).
ISBN: 1-892512-40-8). [38] Y. Zhang, P. Murray-Rust, M.T. Dove, R.C. Glen, H.S. Rzepa, J.A.
[25] L. Blanshard, R. Tyer, M. Calleja, K. Kleese, M.T. Dove. Townsend, S. Tyrrell, J. Wakelin, E.G. Willighagen. JUMBO--An
Environmental molecular processes: management of simulation XML infrastructure for ecience. Proceedings of the UK e-Science All
data and annotation. Proceedings of the UK e-Science All Hands Hands Meeting 2004, pp. 930934 (2004), (ISBN 1-904425-21-6).
Meeting 2004, pp. 637644 (2004), (ISBN 1-904425-21-6). [39] ´
A. Garcia, P. Murray-Rust, J. Wakelin. The use of XML and CML in
[26] S. Wells, M. Alfredsson, J. Bowe, J. Brodholt, R. Bruin, M. Calleja, computational chemistry and physics programs. Proceedings of the
R. Catlow, D.J. Cooke, M.T. Dove, Z. Du, S. Kerisit, N. de Leeuw, UK e-Science All Hands Meeting 2004, pp. 11111114 (2004),
A. Marmier, S. Parker, G.D. Price, B. Smith, H. Spohr, I. Todorov, (ISBN 1-904425-21-6).