MMSN -- Remote Access and Automation Workshop Abstracts

Easter Monday March 28 and Tuesday March 29

The TelePresence Microscopy Collaboratory: Evolving the Collaboratory Paradigm .

Nestor J. Zaluzec Argonne National Laboratory Electron Microscopy Center, Materials Science Division Argonne Il, 60440 USA; Email:zaluzec@aaem.amc.anl.gov

Over the last few years we have made large strides in the adoption of computing and communcation technology to scientific equipment. We have gone from running all our instruments and performing data analysis by "hand" to in some cases complete computer control. But where is this heading, what are the benefits, the pitfalls or areas that need work. Virtual electronic laboratories , or collaboratories and Grids are now being explored in a range of scientific disciplines to exploit this massive investment in computing and communication technology. How can/will TelePresence Collaboratories and Grid techonology change the way we work, the science we do or the way we interact?

For distributed collaboration to be successful in a microscopy/microanalysis (or any discipline for that matter), all of the aspects of the research/education environment must be explored and solutions implemented that match the needs of the research. Deploying resources which donot solve access problems is clearly conterproductive yet more often than not purely IT solutions do not match bench scientists research needs. To achieve the goal of unencumbered collaboration a number of globally relevant issues, in the sense that they apply to all scientific investigations, must be addressed. These include: Persistence Electronic Spaces, Sharable Resources, Sharing Techniques/Protocols, Security/Access Control, Discovery Mechanisms, Transport Protocols, Resource Management and Real World User Interfaces. To effectively implement a collaboratory each of these factors must be understood and an environment created which allows both local and remote individuals or groups to interact in a seamless manner, one in which the visual and audio queues are real time, as appropriate, and in which the "collaboration" is not limited by the physical distance between the participants.

The ANL TelePresence Microscopy Collaboratory (http://tpm.amc.anl.gov), which began operation in 1994 was one of the early attempts of this concept. Using an architecture developed at ANL it is now possible for anyone using the Internet to share resources (expertise, data, and instrumentation) using conventional desktop computers and a modern WWW browser. The key to the optimal implementation of TelePresence Operation, at a given location, is to understand the problems, instrumentation, relative importance and the costs (both in terms of $$ and man-hours) of the various technologies involved and the needs and capabilities of the users. Furthermore, the very nature of what one means by collaboration which varies from discipline to discipline must be defined and each of it's various components considered. Equally important are the sociological issues revolving around on-line resources: be they instruments, data and/or people.

While we have not yet achieved the goal of creating a virtual laboratory which is "as good as being there", significant barriers have been overcome and functional collaboration of remote resources is a reality today. With the record breaking growth we are seeing in computational hardware and software, as well as the solid groundwork being laid by the various collaboratories now being established around the world, it is clear that, in the next 5 to 10 years, the paradigm of virtual collaboratories will become fact and not merely concepts which existed in the mind's eye of a few individuals.

Grid Enabling of the Australian Synchrotron .

Paul Davis, GrangeNet Director, Building 9, Banks St, Yarralumla, ACT 2600. www.grangenet.net

Abstract: In late 2004 a project team visited prospective synchrotron users in five states to discuss their computing, data-storage, network, visualisation and tele-science requirements. The objective was to consolidate the responses and research conclusions into a scoping document for grid-enabling the Australian Synchrotron. This talk will present the results of the consultative process and describe some of the Grid options.

UK Atlas Petabyte Data Store and Data Curation. .

David Corney, Group Leader Data Storage, Atlas Data Store, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire, UK OX11 0QX. www.e-science.clrc.ac.uk/web/projects/data_storage_and_management

The Atlas Petabyte Data Store (APS) at Rutherford Appleton Laboratory in the UK has been in use for over 20 years, and in that time has gone through four major hardware upgrades. Its currently providing a data storage service as a key component of the Tier1 centre for the UK LHC experiment, and to CCLRC as part of the on-going programme to grid enable the facilities: (ISIS, Diamond, and others). It is also providing a data storage and archive service to the wider UK academic community, including a recent new project with BBSRC to provide data archive service to its institutes for ten years. The talk will describe the APS system and the service, including the close collaboration with the Data Management group at CCLRC who provide the increasing popular SRB (Storage Resource Broker) service layer as a powerful interface to gain access to the APS. It will review the APS users, expected future usage, and will highlight the current challenges being faced by the development team over the next few years and summarise how those challenges are being faced.

Some of the data has been on the system for over 20 years. New projects are expecting us to store data on the APS for ten years or more. This requires repeated migration of the data onto the emerging technologies, and is closely associated with the growing need for data curation within the scientific community. The talk will summarise the role of the recently established Digital Curation Centre within the UK and in particular CCLRCs role within the consortium.

e-HTPX Project for Grid based Automation and Remote Access to the Daresbury and Diamond Synchrotrons..

David Meredith and Ronan Keegan, Grid Technology Group / e-HTPX Project, Daresbury Laboratory, Warrington, WA4 4AD, UK. Web: clyde.dl.ac.uk/e-htpx/index.htm

The e-HTPX project centres on the development of an e-Science environment designed to allow structural biologists remote, integrated access via Web-Portal and Grid technologies to Synchrotron facilities and services developed for high-throughput protein crystallography. This includes access to automated beam-line data collection (DNA), HPC data-processing services and access to remote data-transfer/storage facilities. The project encompasses all stages of high-throughput protein crystallography from protein production through to deposition of the final refined model into protein-structure databases. Emphasis is on the unification of all procedures involved in protein structure determination. This will be achieved by the development of an easy-to-use, all-in-one portal interface from which users can initiate, plan, direct and document their experiment either locally or remotely from a desktop computer.

Important concerns for (remote) users of a distributed system include the speed and security of data transferred between remote sites. To achieve security and efficient transfer of data, user-certification, grid technologies (e.g. Globus Toolkit) and data-management systems (e.g. Storage Research Broker - SRB) are currently being employed. Automated techniques are required for speeding-up the data collection and structure solution processes. The goal is to provide users with close to real-time feedback on the quality of their crystal data. To achieve this, we are making use of technologies such as cheap Linux clusters and clustering software (Condor). Automation is also being introduced for the transfer of crystal samples and their meta-data from home laboratories to chosen Synchrotron facilities using a barcode tracking system. To keep track of the experiment as it proceeds, a common data-model allowing for ease of communication between all of the sites involved has been produced. The project is spread across several different sites, including Daresbury Laboratory, York University, the OPPF in Oxford, the EBI in Cambridge and BM14 at the ESRF in Grenoble.

The presentation will also include some of the main points in the following summary of remote access and security options:

Desktop Simulators (e.g. VNC) VNC simulates the desktop of a remote workstation which is viewed directly on a users personal pc. It therefore provides a means for full remote visual interaction and facilitates (remote) execution of GUI tools/programs installed on the host and for computational steering. VNC authenticates the user with username and password. It supports several different platforms including Linux, Windows and Solaris. Firewall restrictions imposed by service-sites may require the client to tunnel through ssh (ssh port - 22, default ports 5900, 5800 for Java version). VNC free version provides no encryption 'over the wire', but is available with the commercial enterprise edition. The Enterprise version supports 2048 RSA keys for the server authentication, 256 character usernames and passwords for client authentication and 128-bit encryption for secure communication. Specific requirements on the remote host include set-up of a user account (for the remote client) and installation of the VNC server software.

Grid Environments Grid provides a means of accessing multiple Grid-enabled host machines and their associated resources/functionality, via both 'generic' interfaces (e.g. GSIssh if supported by the host), or by interfaces that can be customised for a particular service or project. For example, web-portals (e.g. e-HTPX Grid portal) can be used to provide remote access to project-specific computational services (e.g. HPC clusters running Biological codes and Beamline Machines capturing data). Grids provide a means of significantly increasing computational power and potential storage capacity by linking multiple Grid-machines together. The main features of a Grid are centred around:- a) submission and monitoring of computational jobs on remote resources (either via Grid-enabled workstations/tools or portals), and b) high-performance data transfer between multiple hosts (including 2^nd and 3^rd party transfers involving data flow between two remote hosts and/or moving data to/from the remote host and personal workstation). Remote users are authenticated with their personal certificate rather than by passwords, and all transactions are strongly encrypted.

Security and data-confidentiality is of significant importance to Grid. User authentication is handled through the use of personal X.509 v3 digital certificates, issued by Certification Authorities. For international users, CA?s support collaborational-agreements. Use of X.509 certificates ensures user-authentication as photo-id has to be presented by the user before the CA issues a certificate. In addition, 'over-the-wire' transactions are strongly encrypted with the users certificate. Grid portals (e.g. E-HTPX) often adopt the Grid Security Infrastructure (GSI) mechanism and a MyProxy server. A MyProxy server provides a secure online repository for users to store their certificate so that it is accessible from anywhere in the world. The benefit is not just availability (you do not need to take your certificate with you when you travel), but more importantly security, as the MyProxy server will only pass on a subset of your credentials to the requesting Grid-service (this is a time-limited version of your certificate known as a 'proxy'). Your permanent credentials are therefore safe and should never leave your personal workstation. When authenticating to a Grid (e.g. Via a Grid-portal or Grid-enabled work-station), the other Grid-enabled services/hosts that constitute the Grid use the 'proxy' to authenticate the user (rather than re-submitting usernames and passwords). The user can then benefit from single-sign-on-authentication to multiple Grid resources.

Advantages of VNC:

  • Easy to install and setup.
  • No need for client side installation if Java client is used through a web browser.
  • No requirement for further development of data processing software.
  • Graphical interaction with software facilitates computational steering.
  • Authentication and encryption provided with enterprise edition.
  • Support for different platforms, e.g. a Windows client can view a Linux server and vice-versa.

Disadvantages of VNC:

  • Requirement for SSH port or 5900/5800 ports to be open through a site firewall.
  • Not very task specific. A remote user is given access to all the functions of the server machine.
  • Connection between client and server can be potentially slow depending on capacity of network infrastructure between them. Particularly for graphical intensive tasks such as remote viewing of diffraction images.
  • Client does not require certification to access the server.

Advantages of Grid/MyProxy:

  • Authentication and certification of users along with strong communication encryption.
  • Possibility for increased speed for data transfer through the use of GridFTP, e.g. 3rd party transfer of images from beam-line to home labs via a portal interface.
  • Allows for the development of task-specific services.
  • Provides remote access to facilities such as HPC clusters and large data storage servers (can significantly increase computational power/resources).
  • MyProxy allows for delegation of certificate credentials to any machine, allowing users to access the Grid services from anywhere.

Disadvantages of Grid/MyProxy:

  • Longer and more difficult to implement.
  • Difficult to install and setup.
  • Requirement for users to go through the process of gaining a Grid certificate which may make it unattractive to potential users (i.e. 'obstruction at first hurdle').
  • Requires certain ports to opened through site firewalls.
  • Difficult to support visually interactive codes and GUI's (thus job-steering).
Design and Implementation of a Remote Operation System for Ultra High Voltage Electron Microscope for Telescience.

Toyokazu Akiyama and Shinji Shimojo, Cybermedia Center, Osaka University,
web:http://www.cmc.osaka-u.ac.jp

The Research Center for Ultra High Voltage Electron Microscopy (RCUHVEM; www.uhvem.osaka-u.ac.jp) maintains an ultra-high voltage electron microscope (UHVEM) with a 3MeV acceleration voltage. Since a high energy electron beam is used to project the specimen image, it generates X-rays upon contacting microscope column components. In order to prevent X-ray irradiation to observers, the RCUHVEM requires a remote control system. At the same time, the National Center for Microscopy Imaging Research (NCMIR) at the University of California, San Diego (UCSD) applies electron microscopy techniques to neuroscience challenges and develops technologies for computed electron microscopic tomography for use with their microscopes. Since the UHVEM has a very high acceleration voltage, it can observe very thick specimens. NCMIR is interested in observing biological specimens as a 3D volume without slicing and reconstructing their structures. In order to make their experiment comfortable, we have collaborated with the Telescience Project to enable remote operation of the UHVEM.

The NCMIR is developing a Grid based telescience portal for remote control and specimen image processing (https://telescience.ucsd.edu/). Grid middleware technology is being used for image processing and storage; the Ultra High Voltage Electron Microscope (UHVEM) is equipped with high a resolution CCD camera which can output 4k x 4k pixel images. In order to reconstruct a 3D image from a 2D projection images, at least 180 images are required (one image every 2 degrees), and reconstruction is accelerated with Grid parallel processing. Storage Resource Broker (SRB) is used for data and reconstructed image storage, distribution and sharing among researchers.

New remote control software based on Globus Toolkit 3 Web Services is currently under development, including device drivers for the microscope, CCD and other devices. A Java based control client is used to access the GT3 service. Also to be developed is an authentication system for the GT3 control system. The Cybermedia Center is likewise interested in developing a Grid Web Services security based system for Osaka University. This will implement JSR168 and an OpenSAMLbase portal prototype, and it is hoped this can be applied to the telescience environment.

An attractive advantage of a Web Services approach is that it is easy to integrate the implemented function into other applications. For example, the NCMIR is developing a Storage Resource Broker client interface with a Java and GSI base implementation. It enables users to drag and drop data files seamlessly from data grid environment to the other applications, e.g. an image processing application. Although some of the remote operation software like VNC also support interaction with local applications, it is difficult to pass data or parameters from VNC to the other applications because it does not have a standard interface to implement such interactions. A potential disadvantage of using a Web Services or Grid Services approach to remote access is that it requires end-to-end communication from user to devices, with a potentially significant protocol overhead that may limit performance.

In the NCMIR approach, Web Service programs are integrated with Grid Services providing security and convenience with Grid technology. PKI based technologies can now also combine the same kind of security and convenience. Grid technology does not always provide the better solution.

For example, in WS-Security, a Single Sign On standard protocol SAML is proposed by W3C. It provides a more sophisticated SSO mechanism than does GSI used by the Globus community. GSI can produce proxy certificate at any node, and may generate vulnerabilities because of the large amount of resources managed in a Grid environment, which may not be easy to keep secure.

Furthermore, we have PKI based VPN technologies such as SoftEther CA (VPN product we are testing), which can provide seamless access (L2 VPN) to remote environment. Remote resources can be used as resources in local subnet. Such technologies can be used with PKI authentication devices (smart card devices) such as a USB key, or IC card. A user need only insert a USB key to a machine and on input of a PIN number, any resource connected to the network, with authentication provided by the certificate installed in the key device.

As just described, an SSO environment can now be established without Grid technology; suggesting a need to now review the role and relevance of the Grid. The Telescience Project is currently addressing these issues. Interestingly the next version of Globus (GT4) will likely integrate Web Service technologies such as SAML into their implementation.

The Cybermedia Center is currently trying to construct PKI infrastructure for Osaka University, in collaboration with 7 computing centres in Japan; Hokkaido University, Tohoku University, University of Tokyo, Nagoya University, Kyoto University, Kyusyu University. This will incorporate federated SAML authentication such as offered by the Internet2 Shibboleth project. Such technology may also provide a valuable authentication base for remote instrument access.

Methods, Software, and Equipment for High-Throughput Data Collection in Macromolecular Crystallography: the Brookhaven Experience.

R.M. Sweet, D. Schneider, H. Robinson, A. Hiroux, A. Soares, M. Becker, J. Jiang, J.Skinner, R. Buono, and M. Cowan. Biology Department, Brookhaven National Lab., Upton, New York 11973: http://www.px.nsls.bnl.gov.

The PXRR (http://www.px.nsls.bnl.gov) operates six beam lines at the NSLS for macromolecular crystallography (PX). We have defined a new paradigm for use of the synchrotron by structural biologists. Rapid Access, sometimes very rapid, is the norm for most beam line visits. We implemented a mail-in (so-called FedEx) operation: investigators send frozen specimens; we do the work. We provide fluid access to all of our work stations and a consistent look-and-feel among them. It's not unusual for visitors to switch to a more appropriate beam line during a visit, or to operate two at once.

To accomplish this we provide a pool of equipment and personnel, especially overnight PX Operators and computing support. We have a long-standing drive to provide users with easy-to-use tools that let them focus on the crystallographic problem at hand. Software development is a central aspect of this; more recently we are installing ALS-style automounters. Now we are tying the whole process together in a high-throughput mode with our experiment-tracking database, PXDB, which carries information from the initial request for beam time to the final reduced data. All of this is supported by our numerous innovations in the field: the first electronic area-sensitive detector at a beamline in the US, the first graphical user interface for experiment control, on-site data reduction, automatic MAD data collection, web-based remote observation, and more.

Grid and Remote Access for UK National Crystallography Service. .

Mike Hursthouse, EPSRC National Crystallography Service, School of Chemistry, University of Southampton, Highfield, Southampton, SO17 1BJ, UK.
Email: M.B.Hursthouse@soton.ac.uk
Web: http://www.soton.ac.uk/~xservice, http://www.combechem.org/


Ken Meacham, IT Innovation Centre, 2 Venture Road, Chilworth Science Park, Southampton, SO16 7NP.
Email: kem@it-innovation.soton.ac.uk
Web: http://www.it-innovation.soton.ac.uk/

Recent developments in scientific instrumentation have led to an explosion in the rates of data generation and capture that are now possible, and, simultaneously, the demand for such data. In many subjects, this has meant that work-up and interpretation, together with publication or other methods of dissemination, lag seriously behind the generation of the data. This is particularly true in X-ray diffraction studies of chemical structures, and the associated measurement of chemical and physical properties of crystalline compounds. In the Southampton Comb-e-Chem project, a multi-disciplinary project funded by the UK e-Science initiative, we are using the techniques of e-Science and Grid computing to create procedures to address this problem. Our objective is to develop an end-to-end process from the e-Lab to e-Publication, including automation of equipment and analysis, with provision for remote access to experimental equipment and data sharing. Our processes aim to provide support for a wide range of data types, including multimedia and visualization, graphical tools for process capture for data analysis, and utilisation of a statistical approach to the design of experiments. The ultimate aim of the project is to integrate these activities into a Grid environment, with detailed attention to security aspects, and with incorporation of provenance, making the information on the Grid auditable.

The UK National Crystallography Service (NCS) at Southampton is being used as an exemplar for this study. The NCS provides UK academic chemists and crystallographers with access to a full structure determination or data-collection-only service, and currently has over 60 users, with over 90 approved projects. The current annual output is over 2000 data collections and structures. The Grid-enabling objectives for the service have been to create, first of all, a sample and data handling service, featuring:

  • Sample meta-data submission.
  • Sample status tracking (e.g. scheduling of experiments).
  • Acquisition of processed experimental data (e.g. experimental reports, HKL files) and/or derived results (e.g. crystal structures).

Alongside this has been the development of a remote access facility to enable full user participation in the experiment. For this, we are further automating our data collection equipment and data handling software. We are improving methods for secure storage of data and metadata, both raw and enhanced, and we are developing software that can provide new methods for automated interrogation of large datasets, to extract embedded knowledge, and also access to follow-on computational chemistry resources, as Grid software services.

For the NCS Grid Services, we have considered security of the systems and data to be just as important as the functionality. Remote users must have guaranteed access to only their own data and experiments, and they must be fully trusted and authenticated. Within Comb-e-Chem, we have developed sophisticated, secure sign-up procedures whereby users are issued with NCS X.509 certificates. However, a central requirement was for the NCS systems/procedures to be realistic and operationally robust, so major effort has been taken to ensure reliability of a core set of achievable software components, in preference to a broader spectrum of experimental services. We will discuss the trade-off between tight security versus user-friendliness of systems and procedures.

Telepresence Microscopy at CSIRO Minerals. .

Colin. M. MacRae, A. Torpy and N. C. Wilson,
Microbeam Laboratory, CSIRO Minerals, Bayview Avenue, Clayton, VIC.
Corresponding author email: Colin.MacRae@csiro.au
Web:
Nuclear Microprobe:  http://nmp.csiro.au/
CSIRO Exploration and Mining:   http://www.em.csiro.au/
Australian Microscopy Virtual Laboratory: http://www.tpm.csiro.au/

The CSIRO Microbeam Laboratory at Clayton has a range of electron microscopes and serves site based clients, off-site CSIRO clients and external customers. To allow easy access to the electron microscopy facilities we have employed a series of approaches over the last five years including:

  • Web based control/video and note-booking.[1]
  • Java based microscope control/video and microscope control.

In this paper we present the latest TelePresence system developed within the CSIRO Minerals Microbeam Laboratory [2]. This system includes advanced video compression and transmission technology, strong security features, modular and extendible design architecture, and a powerful, easy to use client interface. TelePresence is implemented on two scanning electron microscopes (Hitachi S-5000 FESEM, FEI Quanta FEG-ESEM), and one electron microprobe (JEOL JXA-8900R superprobe).

Using a standard internet connection, remote clients can receive high-quality streaming video as well as data such as spectra and maps, all in real-time (Fig. 1). Some control interfaces are included by which the client may operate instrumentation themselves. However, the system is principally employed as a collaborative tool in which a trained operator is present to control the instrument, and the remote client watches 'over-the-shoulder'. This modus operandi allows the client to better direct the focus of the analysis as it provides them with immediate results, data interpretation and expert advice. Thus far, this system has both enhanced the collaboration with our existing clients, and also attracted new work with distant customers.

A key feature of the system is the provision of high frame-rate video with low-latency. Frame sizes of 640 480 pixels at 25 frames per second have been achieved over standard cable modems, with bit-rates less than 2Mbits/s. Video is compressed using the MPEG-4 standard, with a video compression-decompression latency of less than 200ms. Using this protocol a key frame is initially sent followed by changes. While the transmitted data rate spikes during key frame transmission overall data transmission rates are significantly reduced. Alternatively, compression standards are being trialled which enable the end user to customise the video quality and trade-off frame rate when required. Originally, "transmission control protocol" (TCP) was employed to transmit video but transmission latency was found to be high, giving up to several seconds delay. This latency has been minimised through the use of the connectionless "user datagram protocol" (UDP), which does not require a return acknowledgement for each packet transmitted. A standard TCP/IP socket is used to transmit all other data and is encrypted using the 256-bit 'Rijndael' cipher otherwise known as the NIST advanced encryption standard [3]. This cipher guarantees that all data is entirely confidential when transmitted over the internet. Audio is transmitted over the telephone system.

Figure 1. Telepresence user interface showing live video and streaming energy dispersive spectrometry (EDS) from the x-ray detector.

Generic and inexpensive hardware components have been used throughout the system as insurance against discontinued availability of essential components. TelePresence 'servers' consist of a 2GHz+ desktop PC, Microsoft XP, with a low-cost video capture card, and the client need only have a 500MHz+ PC with a good network connection (preferably 1Mbit/s or better). The use of 'low end' components has not adversely affected the stability of the system, as multiple parallel client connections have been maintained for several days without disruption.
References:

[1] Zaluzec N., TPM System Description,
www.amc.anl.gov/Docs/ANL/TPM/TPMHomePage.html
[2] A. Torpy et al, AMAS VIII Symposium, Feb. 2005, pp 85
[3] http://csrc.nist.gov/cryptotoolkit/aes/rijndal

FIP: An Automated Beamline for Protein Crystallography at the ESRF. .

Jacquamet L., Bertoni A., Borel F., Charrault P., Fieulaine S.,
Israel-Gouy P., Kahn,R., Joly J., Ohana J., Pirocchi M.,
Serre, L and Ferrer J. L., Institute of Structural Biology,
41 rue Jules Horowitz, 38027 Grenoble cedex, FRANCE .

Crystallography is a method of choice for obtaining structural information on biological macromolecules. To meet growing demand, FIP (French beamline for the Investigation of Proteins) was constructed at the ESRF (European Synchrotron Radiation Facility). Currently managed by the synchrotron group of the Institute of Structural Biology (IBS), this beamline is exclusively dedicated to the crystallography of biological macromolecules. The different optical elements of FIP deliver a focused X-ray beam of 300 × 300 µm2 with an energy tunable from 6.5 to 17 keV for anomalous experiments.

An X-ray diffraction experiment on a synchrotron beamline includes four steps;

  • setting of the X-ray beam energy and optimization of its intensity;
  • mounting and centering of the crystal on the diffractometer;
  • recording of the diffraction pattern;
  • analysis of the data recorded.

To increase the beamline yield and optimize the experiments, each of these steps has been automated. The energy adjustment and optimization are completely automated [1]. Similarly, the positioning of the crystal is ensured by CATS (Cryogenic Automated Transfer System) [2]. The automation of the data recording and processing with the ADP (Automated Data Processing) [3] software has also been achieved.

The FIP team pushes automation even further and use the CATS robot to replace the diffractometer. Ultimately the robot should be able not only to position the crystals in the X-ray beam, but also to rotate the crystal correctly for data recording [4].

In addition, this robot offers the possibility of analyzing crystals directly as they grow in drops inside crystallization plates [5]. Thus the nature of such crystals, protein or salt, can be determined without disturbing the drop. This distinction, which is vital in crystal growing, is facilitated by this system. Furthermore, in favorable conditions, the system is able to record complete data sets. The last development60% of the beamline time is assigned to the French scientific community and 7% to the FNRS (Belgian national scientific research fund) by decision of a programme committee. 10% of the time assigned to the French community can also be sold to industrial operators. Lastly, the FIP team itself is a user of beamline time along with the other French research teams, and users can thus benefit from its experience in the use of synchrotron radiation. The remaining 33% is shared out among international users by the ESRF. is the remote control of FIP: CATS EYES via a fast Internet access [6]. Depending on the Internet access speed at the meeting, a live control of FIP will be realized to demonstrate the different automations.

Since the beamline was launched in 1999, more than 140 crystal structures have been resolved. The ongoing instrumental developments should further improve yield, and also allow projects in which a large number of samples have first to be tested before a satisfactory one can be found.

References:

  1. Roth, M et al. Acta Crystallographica, (2002) D58, 805-814.
  2. Ohana, J. et al. Journal of Applied Crystallography, (2004) 37, 72-77.
  3. Ferrer, J. L. Acta Crystallographica, (2001) D57, 1752-1753.
  4. Jacquamet, L. et al. Acta Crystallographica, (2005) D60, 888-894.
  5. Jacquamet, L. et al. Structure, (2005) 12, 1219-1225.
  6. Joly, J. et al. in preparation.

CIMA: Scientific Instruments as First Class Members of the Grid. .

Rick McMullen
Knowledge Acquisition and Projection Laboratory, Pervasive Technology Labs.
Indiana University.
Ken Chiu
Department of Computer Science, Thomas J. Watson School of Engineering
and Applied Science, State University of New York at Binghamton.

An aspect of Grid computing (a critical enabling component of the Cyber infrastructure envisioned by the NSF) that has not yet been well developed is the integration of scientific instruments into a Grid computing environment. Instruments are still largely 'off-line' as far as downstream software analytical components are concerned and instruments are not at all first class members of the Grid with respect to location, allocation, scheduling, and access control. This is a serious problem as three issues continue to grow in importance in research;

  • investments in geographically extended (e.g., international) collaborations organized around large shared instrument resources;
  • increasingly real-time use of instruments by remote researchers both for 'first look' activities and pipelined data acquisition and reduction;
  • sensor networks with hundreds to thousands of nodes being deployed.

The Common Instrument Middleware Architecture (CIMA) project aims at 'Grid enabling' instruments as real-time data sources to improve accessibility of instruments and to facilitate their integration into the Grid. A primary challenge addressed by this research program is to develop a generalized approach to instrument middleware that allows existing and new instruments to be integrated into Grid computing environments.

CIMA middleware is based on current Grid implementation standards, and is accessible through platform independent standards such as Web Services. In keeping with the design goal of general applicability, CIMA interfaces are being developed for a variety of instrument and controller types from large shared facilities (synchrotron X-ray sources, robotic optical telescopes) to small embedded industrial controllers and sensor nets and micro-sensor packages such as MOTEs, developed for the DARPA SmartDust program.

Other issues being explored include extending the accessibility of instruments to new classes of users (e.g. student remote access to advanced materials science facilities at national labs), use of instruments by software agents, and increasing the longevity, flexibility and durability of software systems for instruments and sensors.

This presentation will present an overview and status of the CIMA instrument middleware project.

Crystallography and Tomography Automation and Remote Access at the Advanced Light Source.

Gerry McDermott, Advanced Light Source, Berkeley.

The presentation will cover the automation and remote access requirements of two distinct experiments at the Advanced Light Source in Berkeley. Protein crystallography is a mature, well established field, where solid progress has been made in reducing the need for researchers to physically be present during data collection. Details of the instrumentation that make this possible will be presented. X-ray Tomography, on biological molecules, is an emerging technology. Currently, we are constructing a new X-ray microscope, dedicated to biology, at the ALS. Details of this new experiment, and our plans for implementing remote access methods will also be discussed.

Grid Computing and e-Science at ISIS.

W.I.F. David, K. Shankland, T.A.N. Griffin and A.J. Markvardsen
ISIS Facility, Rutherford Appleton Laboratory, Chilton, OX11 0QX, UK

The ISIS Facility at the CCLRC Rutherford Appleton Laboratory is currently working closely with the CCLRC e-Science centre in order to ensure that emerging computing technologies (including, but not limited to, 'Grid-type' technologies) are fully utilised at the facility where appropriate. In this regard, there are a number of noteworthy initiatives:

  • ICAT: The ISIS Catalogue (ICAT) addresses the problem of search and retrieval of ISIS data, starting with our back catalogue of over 20 years of neutron scattering data. Developed around an Oracle database and accessed by webservices, ICAT will the definitive source for ISIS data. The system will be fully integrated into the larger scheme ofthe CCLRC Data Portal, which will permit searches across multiple data catalogues (e.g. neutron and X-ray) via a web/web-services interface.
  • ISIS Proposals system: As of April 2004, all experimental proposals at ISIS are now dealt with electronically via the WWW. Developed around an Oracle database, the new system is the definitive source for ISIS proposal data. The system is fully compatible with ICAT, thus allowing linking of experimental data back to proposals. Both the proposals system and ICAT access a new common user database that replaces numerous existing user databases.
  • eVe: The excitations visualisation environment system aims to provide users of new ISIS instrumentation (MAPS,MERLIN) with access to state-of-the-art visualisation facilities for examining and interacting with their experimental data. This involves developing back-end visualisation and computational services that can hook into existing analysis software in order to make the examination of very large, complex datasets tractable on relatively modest front-end hardware.
  • SCARF: The 'super computer at research facilities' is a 256 processor Beowulf-style cluster purchased by thee-Science department specifically for use by CCLRC facilities. Run along the same lines as the UK's National Grid Service, it will act as a powerful compute resource for data analysis.
  • Access Grid: We have recently installed an Access Grid node on the PEARL beamline with a view to exploring the possibilities of remote supervision and/or collaboration on experiments. Providing a far superior experience to simple telephone calls, we anticipate that this experiment will see AG nodes migrate onto other beamlines in the not too distant future.

At a departmental level, we have also been exploring the use of distributed computing for compute-intensive jobs. We shall consider the scenario where a large compute job (one that would take several days to execute on a conventional single personal computer) consists of a series of shorter jobs (each taking typically a few hours to execute) that can executed independently of one another. Hence the short jobs can, in principle, be distributed to various computers for execution and then the results collated. The Grid MP system (www.ud.com) of distributed computing is ideally suited to the routine running of such compute intensive tasks.

In the Grid MP system, there are two key hardware components: the server computers and the client computers that are registered with the server. The servers comprise two dedicated dual-processor PCs running Redhat Linux and they act as the focus for the splitting and distribution of the large compute jobs. The server also keeps track of all jobs and client computers that are part of the 'grid' system. The client computers are standard desktop PCs located within an organisationand they run a small piece of United Devices software called an agent which communicates with the server. The agent has the capability to execute computer programs on the clients in such a way that the people sitting at the client computers are not aware that these programs are in fact running.

Thus the following sequence of events typifies a Grid MP job:

  • A computer program that carries out a series of CPU-intensive calculations (i.e. a large job) is submitted to the server computer from the desktop PC of the person who wishes to obtain the results from the calculations. The submit program splits the job into its constituent smaller jobs.
  • The server distributes these smaller jobs to client machines that are registered with the server i.e. desktop PCs that are running the United Devices 'agent' software.
  • On each client, the agent executes a small job at the lowest operating system priority and utilises only CPU cycles that the client computer is not otherwise using i.e. the user will not notice that their CPU is in fact being used.
  • When a small job finishes executing on a client computer, the agent returns the results of the calculation to the server for collation with the results gleaned from other clients.
  • When all the smaller jobs are complete, the person who submitted the job can retrieve the results from the server onto their desktop PC.

Such a system can be applied to any number of scenarios encountered at ISIS, such as molecular dynamics or RMC simulations, neutron instrument simulations, image generation etc. Here, we will consider the use of global optimisation methods that have increased the size and complexity of molecular organic structures that can now be solved directly from powder diffraction data. These methods require many repeat runs to be performed in order to confirm the location of the global minimum in parameter space; this is particularly true of very complex structures, where success rates in locating the minimum may fall to only a few percent. Fortunately, these multiple runs can be performed independently of each other and as such, they are ideally suited to the notion of grid-type computing. We have recently adapted the DASH structure determination package to run under Grid MP and the current setup allows up to eighty DASH runs to be executed in parallel on existing desktop resources. This presentation will focus on results obtained using both DASH and the HMC structure determination programs. These show not only impressive performance gains but also indicate that new computational routes that were previously closed to us due to their compute-intensive requirements are now open.

X-Ray Tomography System, Automation and Remote Access at Beamline 2BM of the Advanced Photon Source.

Francesco De Carlo and Brian Tieman
Advanced Photon Source, Argonne National Laboratory, Argonne, IL 60439, USA.

X-ray microtomography is becoming the tool of choice for three-dimensional (3D) imaging of thick structures at the 1-10 µm scale. The fast microtomography system developed at beamline 2-BM of the Advanced Photon Source (APS) offers near video-rate acquisition of tomographic data, pipelined processing and 3D visualization combined with fully automated and remotely controlled capability. At its maximum throughput, the system can image hundreds of specimens a day.

The entire instrument, including the tomography setup, beamline and a dedicated 32-node computer cluster for data analysis, is remotely accessible via Access Grid technology giving a user full remote control of every aspect of the experiment.

In this talk we will describe the system and the implementation of the Access Grid technology at beamline 2BM of the Advanced Photon Source.

Remote Access Tools for Macromolecular Crystallography Experiments at SSRL. Mike Soltis
Macromolecular Crystallography Group, Stanford Synchrotron Radiation Laboratory, MS 99, 2575 Sand Hill Road, Menlo Park, CA 94025 U.S.A. USA.
Web: http://smb.slac.stanford.edu/public/index.shtml

The macromolecular crystallography beam lines at SSRL have evolved into an efficient and productive X-ray diffraction data acquisition and processing resource for determining macromolecular structures. The traditionally heroic synchrotron experiment is now a technically robust and elegant research tool for the biomedical research community. With the implementation of automation and the development of collaborative and remote access tools, the resource can be successfully utilized by scientists around the world. Enabling developments in instrument control, automation and remote access tools for streamlined experimental design and execution including the interface with the home laboratory will be presented.

Winners and Losers - Ranking Crystals from Diffraction Images.
A.R. Criswell, R. Bolotovsky, T. Niemeyer, R. Athay, J.W. Pflugrath.
Rigaku/MSC, Inc., 9009 New Trails Dr., The Woodlands, TX 77381.

With the advent of automation, the frequency of crystal screening has increased. As a result, many home labs and beamlines have developed robotic systems for high-throughput evaluation of crystal samples. Most of these systems still depend on human intervention to judge the quality of crystal samples. An automated system ideally should provide methods to identify desirable samples. Ultimately, the best judge of crystal quality is the statistical evaluation of the integrated intensities, reported as the Rmerge and I/sig(I) for the data set.

Because data screening results in much fewer reflections, and few symmetry mates, it is more difficult to evaluate fully the crystal quality. Crystallographers are left to make an estimated judgment of the ultimate data quality by a visual inspection followed by preliminary lattice evaluation.

Previous evaluations of crystal quality scoring have included several methods. One group utilizes a cascade-correlation neural network to monitor and predict diffraction quality based on a training set of 711 diffraction patterns collected at the National Synchrotron Light Source.[1] Another method uses a heuristic measure of crystal quality based upon a calculation which takes into account the mosaicity, refinement residuals, and diffraction limits.[2] Neither of these methods, however, takes into account the inherent anisotropy displayed by a large number of protein crystal samples. Moreover, neither method allows users to fine tune crystal quality calculations or 'pick' their favorite crystal attributes.

We have expanded d*TREK, in which automatic processing and data collection strategy are both already implemented, to include an algorithm to evaluate the quality of diffraction images and assign a rank per sample. This ranking software evaluates images in terms of several rules and calculates an award or penalty for each rule. The awards and penalties are then summed and updated on a per sample basis. Samples can then be ranked according to these values, and data collected in descending rank order. In general, these rules include the number of reflections per resolution shell, the I/s(I) of reflections per shell, the spot sharpness, the presence of sharp or diffuse rings, sample mosaicity, refinement statistics, and the number of accepted reflections from refinement. This talk will address the utility these rules, their usefulness in the evaluation of diffraction images, and those rules which seem most important for ranking of crystals. References:
[1] Berntson, A., Stojanoff, V., and Takai, H. (2003). J. Synchrotron Rad., 10: 445-499.
[2] Zhang, Z., Sauter, N.K., van den Berdem, H., Snell, G.P. & Deacon, A.M. (2004) In preparation.

Remote Access for NANO-MNRF Instrumentation.
Dr Peter Hines* and Duncan Waddell#.
* Australian Key Centre for Microscopy and Microanalysis, The University of Sydney.
# Centre for Microscopy and Microanalysis, The University of Queensland.

The NANO-MNRF provides a national network of world-class imaging and analysis equipment and expertise for Australian research. Remote access has been identified as an important mechanism for promoting community outreach, enabling vendor support, extending teaching activities and enabling research collaboration.

Remote access can be divided into two categories - active (allowing remote control) and passive (allowing remote viewing of the experiment in progress). In practice there are many variants depending on the expertise of the remote party and the level of trust between parties. Other factors involved in establishing telepresence services are network bandwidth, computing environments and the nature of the instrument.

As such, there is no one-size-fits-all approach to telemicroscopy. The authors will present their experiences in serving remote microscopy to environments ranging from schools in the PNG highlands to technique courses in the room next door.

Software Development for the Computational Grid.
Prof. David Abramson.
Computer Science, Monash University,Clayton, Victoria.

In this presentation I will discuss the software hierarchy and middleware layers being proposed and developed for the Grid. In particular, I will highlight some locally designed tools that make it easy to develop Grid applications in particular niche domains. The Nimrod family of tools support robust science and engineering by distributing parametric and search applications across the Grid. The GriddLeS system assists users in porting legacy codes to the Grid, and facilitates the coupling and distribution of computational models.

New Software and Hardware Tools for Remote Crystallography.
R. Durst, M. Ruf, M. Benning, B. Michell, G. Wachter, J. Kaercher, S. Leo, B. Lancaster.
Bruker AXS Inc., Madison, WI, USA.

The new Proteum II software suite exploits a web-based, client-server architecture to allow remote instrument control. Together with the BruNo automated sample mounting robot, these new tools allow a remote user to automatically mount and align a sample for either screening or complete data acquisition and structural determination. This feature is currently being exploited by a number of smaller colleges to form consortia to efficiently share a single intrument at a central site.

The Molecular and Materials Structure Network is funded by the Australian Research Council.

ARC -- Australian Research Council