Download Theory of Data Life Cycle and more Lab Reports Life Sciences in PDF only on Docsity!
Data Life Cycle Labs
A New Concept to Support Data-Intensive Science
Jos van Wezel, Achim Streit, Christopher Jung,
Rainer Stotzka, Silke Halstenberg, Fabian Rigoll,
Ariel Garcia, Andreas Heiss
Karlsruhe Institute of Technology (KIT) Karlsruhe, Germany Achim.Streit@kit.edu
Martin Gasthuber
Deutsches Elektronen-Synchrotron (DESY) Hamburg, Germany Martin.Gasthuber@desy.de
Kilian Schwarz
GSI Helmholtzzentrum für Schwerionenforschung GmbH Darmstadt, Germany k.schwarz@gsi.de
André Giesler
Forschungszentrum Jülich (FZJ) Jülich, Germany a.giesler@fz-juelich.de
Abstract —In many sciences the increasing amounts of data are reaching the limit of established data handling and processing. With four large research centers of the German Helmholtz association the Large Scale Data Management and Analysis (LSDMA) project supports an initial set of scientific projects, initiatives and instruments to organize and efficiently analyze the increasing amount of data produced in modern science. LSDMA bridges the gap between data production and data analysis using a novel approach by combining specific community support and generic, cross community development. In the Data Life Cycle Labs (DLCL) experts from the data domain work closely with scientific groups of selected research domains in joint R&D where community-specific data life cycles are iteratively optimized, data and meta-data formats are defined and standardized, simple access and use is established as well as data and scientific insights are preserved in long-term and open accessible archives.
Keywords: data management, data life cycle, data intensive computing, data analysis, data exploration, LSDMA, support, data infrastructure
I. INTRODUCTION Today data is knowledge – data exploration has become the 4th pillar in modern science besides experiment, theory, and simulation as postulated by Jim Gray in 2007 [1]. Rapidly increasing data rates in experiments, measurements and simulation are limiting the speed of scientific production in various research communities and the gap between the generated data and data entering the data life cycle (cf. Fig1) is widening. By providing high performance data management components, analysis tools, computing resources, storage and services it is possible to address this challenge but the realization of a data intensive infrastructure at institutes and universities is usually time consuming and always expensive. The introduced “Large Scale Data Management and Analysis” (LSDMA) project extends the services for research of the Helmholtz Association of research centers in Germany with community specific Data Life Cycle Laboratories (DLCL). The
DLCLs are complemented with a Data Services Integration Team (DSIT) which provides generic technologies and services for multi-community use based on research and development in the areas of data management, data access and security, storage technologies and data preservation.
The LSDMA project initiated at the Karlsruhe Institute of Technology (KIT), builds on the familiarity with supporting local scientists at a computer center, the knowledge of running the Grid Computing Centre Karlsruhe (GridKa) [2] as the German Tier 1 hub in the World Wide LHC Computing infrastructure [3], the Large Scale Data Facility (LSDF) [4] and the experience with the very successful Simulation Labs [5] that specialize at supporting HPC users.
Figure 1. The scientific data life cycle
The project partners from the Helmholtz Association have an equal long standing experience in large data and computing support in various communities. LSDMA brings together the combined expertise from all partners as well as the data intensive research communities that are local at their institutes.
II. LARGE SCALE DATA IN MODERN SCIENCE Modern science and scientific computing is about data. In the process from collecting data till publication, the data has been moved, aggregated, selected, visualized and analyzed. In view of the ever increasing amounts of data this process must be organized and structured. Data management is the organization and structuration of the data life cycle which will allow faster results and dependable long term references. The data life cycle does not stop after publication. Without a suitable data organization or infrastructure (e.g. with persistent identifiers) data can no longer be found and is essentially lost. Data has become a source of knowledge in itself [6]. By combining, overlaying, mining, visualizing and application of various analysis methods the scientific process is now also driven by the content of already existing data collected in experiments and measurements. Data from every section of the data life cycle, including the raw measurement or simulations and data from rare and unique events, has become a valuable good that needs to be preserved and is lost if not properly managed.
Common in many experimental sciences are the vast requirements for large scale storage, data management and analysis resources. The actual requirements of the communities must be examined in detail because we have found a large variation of data management awareness and experience across communities in our discussions with them. Some scientific disciplines already have local solutions and are investigating e.g. the possibility of data sharing within the community whereas others are struggling with basic data handling and have only recently begun to organize their unstructured data.
A. Data and analysis in climate and environment
1) Atmospheric research Our understanding of the atmosphere is fundamentally limited by the sparseness of highly resolved regional and global observations. Data from remote sensing instruments like GLORIA [7] and MIPAS [8] on the ENVISAT satellite [9] promises to improve this situation substantially. Both instruments require data storage capacities of several hundred Terabytes per year (raw and processed data). Further projects and instruments are on the horizon, requiring similar amounts of data storage and data management and analysis techniques. The raw data consist of interferograms with a very high spectral resolution that are converted in multiple steps resulting in many (> 10000) small files per day. During processing the access of such a great amount of small files is the limiting factor compared to the computing times for analyzing the data. Due to the irreplaceability and preciousness of the data, they must be archived for several decades
2) Climate modeling Climate model data from HPC simulations is being used worldwide by a growing number of users. The amount of new data ranges from 10 to 100 PB per year in international
modeling activities (e.g. CMIP5 [10] and CORDEX [11]) which are archived and disseminated in globally distributed data federations. In recent years data applications are changing from community specific to interdisciplinary applications. Data federations have been established and the scientific research is treating more and more coupled and interdisciplinary questions like adaptation and mitigation. After annotation and quality control, around 25% of the climate model raw data move into long-term data archives. These model data will not be reduced anymore because usually detailed use cases are unknown in advance and therefore it is not predictable which outcome will be necessary for further investigations. After passing quality control the data is freely available and subsequent access is very unpredictable. Worldwide available archives for climate data must have ample capacity and must assure long term availability as well as related data curation services [12].
Data discovery, efficient data access, optimizing the processing workflows, security and quality management are only several examples which have to be considered for optimizing the data life cycle of climate model data. A fast und secure data access and transfer is also an essential requirement of the community for an interdisciplinary use of the archived data in international federations.
B. Data and analysis in energy research Data intensive computing in energy research comprises many topics that are driven by public interest and reflects the prominent international role Germany plays in the field of renewable energy. Applications at KIT focus on the area of energy generation, intelligent distribution, energy storage and intelligent management. Very large computing requirements in energy research originate from fusion research. In preparation of the international ITER [14] cascades of multi-scalar simulations are run to improve the fusion physics model and the condition predictions of the running reactor. The ITER project also uses distributed computing which implies the need for high speed and reliable data exchange between supercomputers. The experiments such as the large scale parameter studies of plasma analysis must be able to store and track the intermediate and final result of the computations as well as the associated meta-data in which the simulation conditions are kept for future reference. Data must be available for decades in the future since the ITER project is expecting its first demonstrations not earlier than 2040.
In the domain of energy distribution across a country, profiling and analysis of the consumption of electric energy using customer data to optimize the construction and use of the electric power grid is state of the art and often referred to as smart grids [15]. Originally to optimize the efficiency and reduce wear of power generation plants smart grids are increasingly expanded with smart metering in households. The steepness of the power curves is influenced by intelligently monitoring, predicting and controlling the consumption in the power grid and of the power generation in classic plants and increasingly, from alternative sources i.e. solar and wind. To be effective and efficient the analysis results of large scale data acquisition and simulation must be available quickly and in high detail. The amount of data to be stored and analyzed from
The excellence cluster CellNetworks of the University of Heidelberg [24] addresses, among other, biomedical questions in the area of structural biology and structural cell biology. Contributing to an array of scale invariant imaging techniques research in cryo electron microscopy investigates 3D reconstruction of electron microscopic tomogram images using e.g. filtered backprojection and least squares solution of backprojection matrix [25]. The workflow produces almost 300 images with 18 GB data per tomogram. The setup is able to run 40 to 60 tomograms per day using 2 microscopes. 2D image analysis and 3D reconstruction of this data is only possible with a highly automated processing workflow and by moving part of the computation to special purpose nodes using GPUs.
Second generation single plane illumination microscopes (SPIM) [26] are able to reach very high resolutions of 3D images in only 30 seconds recording time. This technology allows recording over a long period collecting information about the influence of certain toxic environments to the development of a living zebra fish embryo. An observation time of 10 hrs results in datasets of approximately 7 TB which have to be transferred to the data facility, stored and analyzed. Analysis contains complex image processing algorithms like registration, segmentation and automatic feature extraction. To avoid bottlenecks resulting in huge delays in data analysis, challenges in high performance data management and data intensive computing have to be solved.
E. Data and analysis in future large scale physics. The FAIR project The Helmholtzzentrum für Schwerionenforschung GmbH (GSI) [27] operates a large and in many aspects unique accelerator facility. Centered on GSI, an international structure named FAIR (Facility for Antiprotons and Ion Research) [28] will evolve. FAIR will generate beams of a previously unparalleled intensity and quality. In the final design FAIR consists of eight ring colliders with up to 1,100 meters in circumference, two linear accelerators and about 3.5 kilometer beam control tubes. The existing GSI accelerators serve as an injector. About 3000 scientists and engineers from more than 40 countries are already involved in the planning and development of the facility and its experiments. FAIR will support a wide variety of science cases: extreme states of matter using heavy ions (CBM), nuclear structure- and astrophysics (NUSTAR), hadron physics with antiprotons (PANDA), atomic and plasma physics as well as biological and material sciences (APPA).
The high beam intensities at FAIR constitute various challenges: very efficient accelerators, remote handling in activated areas, novel methods of cooling for the detectors, progress in collaborative computing and synergies between the various fields of research especially concerning the interaction with industry. The first beam is expected to be in 2018. This constitutes a crucial milestone in computing.
For two of the research topics of FAIR (CBM and PANDA) characteristics of the computing and data analysis are similar to the LHC at CERN that relies on trigger systems, but need further extensions. Trigger systems were necessary due to real-world limitations in data transport and processing bandwidth. At FAIR a novel triggerless detector read-out will
be implemented, without conventional first-level hardware triggers, relying exclusively on event filters. This approach is much more flexible and adaptable to yet unforeseeable needs because the full detector information is available even in the first decision stage. Complex filtering and flexibility are the key enabling factors for the experiments at FAIR. So at FAIR the classical separation between data acquisition, trigger, and off-line processing is merging into a single, hierarchical data processing system, handling an initial data stream exceeding 1 TB/sec. The first layer of the system constitutes the first level event selector (FLES). The FLES implements a combination of specialized processing elements such as GPUs, CELL or FPGAs in combination with standard HPC clusters comprising compute nodes coupled by an efficient high-speed network. After the FLES the data stream fans out into an archival system and into the next distributed processing layers, which can be off-site. Currently large-scale research infrastructures like e.g. LHC at CERN rely on on-site tier-0 data centers which perform the first-level processing after the last trigger step. Subsequently a considerably reduced amount of data is analyzed off-site in downstream data centers. In contrast to that FAIR will make use of a novel modular data processing paradigm using multi- site load balancing data centers. Several sites in the surrounding of GSI will be connected with a high-speed Metropolitan Area Network via fibre link allowing the off- loading of processing between the sites. Multi-site user access and corresponding security aspects are currently being investigated within the EU project CRISP [29]. That combined tier-0/1 system will be integrated in an international Grid/Cloud infrastructure. The FAIRGrid distributed computing infrastructure is currently implemented as two separate entities: PandaGrid [30] and CBMGrid. FAIRGrid uses the AliEn middleware [31], as developed by the ALICE experiment. The grid monitoring and data supervision are done via MonALISA. The basis of the software for simulation, reconstruction, and data analysis for the large FAIR experiments is FairRoot [32]. In order to meet peak demands for computing, it may be necessary to offload some of the computing tasks to public or community clouds. Thus, it becomes important to deploy and operate an infrastructure to compute these tasks on virtual machines in a quick and scalable way.
The exact amount of computing, storage and archiving required for FAIR depends on many factors but is certainly beyond the capacity of a single computer center. The required resources are dominated by CBM and PANDA. Current estimates for the sum of all experiments are 200.000 cores and 30 PB storage space for the first year of data taking. The load balancing distributed infrastructure must be developed as well as the seamless integration with existing grid and cloud infrastructures. The re-use of components from existing running high energy physics experiments poses specific challenges for data management and distributed computing that have to be solved within a few years.
III. R ELATED A CTIVITIES AND INFRASTRUCTURES
A. GridKa
In 2002, the former Research Centre Karlsruhe (now KIT) started GridKa [2], providing compute power and storage for High Energy Physics experiments. Many physics experiments, e.g. CDF, D0, Babar and Compass, used GridKa for production while the LHC experiment collaborations already started to simulate their experiments and detectors. During this time, GridKa was participating in the EU-DataGrid project and later the LCG-1 and LCG-2 Grid projects.
In the following years the number of nodes and storage capacity was increased yearly to accommodate requirements of the particle physics community during which the LHC experiments developed and tested their distributed computing models. Numerous “Data and Service Challenges” have been performed to test these models and demonstrate the capabilities of the various participating Grid centers. In 2008, the Worldwide LHC Computing Grid [3] achieved the required target data rates and was able to demonstrate its readiness to store, distribute and compute the data to be produced by the LHC.
GridKa is one of 11 Tier-1 centers of the Worldwide LHC Computing Grid and one of four centers providing services and resources for all four of the large LHC experiments. It currently provides 130000 HEPSPEC’06 [33], approx. 11 PB of disk space and 17 PB of tape space. The disk and tape space is provisioned by several Grid enabled dCache [34] and XROOTD [35] storage systems, accessible from the wide area networks, capable of delivering I/O rates of several Gigabytes/s. The gLite File Transfer Service [36], controlling wide area data traffic to and from other Grid centers as well as workload management systems, information systems and file catalogues are provided as high-available Grid services.
B. HPC Simulation Labs
At Karlsruhe Institute of Technology (KIT) and the Forschungszentrum Jülich (FZJ) Simulation Laboratories (SimLabs) for different research communities have been established. For example, the SimLab Energy at KIT focuses on models of chemical processes and of flows e.g. of hydrogen combustion and neutron physics, while the SimLab NanoMikro at KIT targets the simulation of nanostructured materials e.g. with density function theory. The other SimLabs at KIT are the SimLab Elementary and Astro-Particles and Climate and Environment. At FZJ the SimLabs Biology, Molecular Systems, Plasma Physics and Climate Science exist – the later one in close collaboration with the resp. SimLab at KIT.
A SimLab [5] is a community-oriented research and support group acting as interface between the supercomputing facilities and the university groups, institutes and communities. Its mission is to provide high level support in utilizing HPC facilities, especially at FZJ or KIT, by porting model systems to supercomputers and subsequently optimizing the performance of these models by enhancing their parallel scalability, load balancing and data access. The focus is support for High Performance Computing (HPC), but a SimLab also performs its own research together with the scientific
communities. This work is typically done in close cooperation with scientific institutes through common projects.
The relationship between SimLabs and Data Life Cycle Labs in LSDMA is that in both instruments IT experts work in close collaboration with domain scientists. In the case of SimLabs on the particular challenges of enabling scientific communities to make efficient use of supercomputers for their simulation codes – in particular considering multi/many-core processors and Peta/Exaflop/s system performance.
C. LSDF The Large Scale Data Facility (LSDF) [4] was started at the end of 2009 at KIT with the aim of addressing the data management and processing requirements of several forthcoming data intensive experiments [37]. In particular, the High Throughput Microscopy experiments used for biological studies of zebra fishes and the Tomography beamline of the synchrotron radiation source ANKA, were planning to collect data amounting to several Petabytes in the coming years. Providing not only the storage capacity, but also the data management and analysis services was identified as a critical aspect where a unified Campus-wide approach would profit from many synergies, and relieve the communities from the administering a IT Infrastructure. Soon after the project started many scientific groups on the campus expressed their interest and joined. Currently the LSDF is providing storage through the GridFTP, NFS, and CIFS protocols [37], as well as an iRODS Datagrid [38]. The experimental data benefits from data management functionality if ingested with the tools developed within the project. Additionally, the data analysis can be carried out by means of the project's Cloud or Hadoop services [39].
A close collaboration of LSDMA and LSDF is natural, as LSDMA builds to a large extend on services and infrastructure such as LSDF provides. Currently, the LSDF is hosting around 1 PB of data and provides Hadoop and Cloud resources to scientific experiments from 17 KIT institutes. The number of scientists using these services lies in the range 80-100.
D. EUDAT The European Commission funds the European Data Infrastructure (EUDAT) [40] aiming at establishing a cost- efficient and high-quality collaborative data infrastructure throughout Europe. It addresses the needs and requirements of various scientific disciplines both in terms of capacity and capability. 25 partners from 13 countries focus on the development and operation of services for federating existing resources, which are provided by the EUDAT partners. Both FZJ and KIT are partners in LSDMA and EUDAT. Further services address common, baseline data services for a variety of communities, long-term archival with bit-stream persistence and integrity for a long-term and distinct referencing of data, proper up/download services for users as well as a federated security infrastructure for authentication and authorization. A close collaboration of LSDMA and EUDAT is foreseen in which national services and processes are aligned with those on European level in order to enable scientific communities and boost the pan-European data-driven scientific discovery.
access, improved data aggregation and proper interfaces for platform integration.
All R&D in LSDMA is based on the communities' requirements for an optimized data life cycle. As this optimal data life cycle evolves with time, R&D in LSDMA has to adjust swiftly, dropping the development of specific software tools and replacing them with other tools.
C. Time Line
As the needs and requirements by the communities change over time, two major reports are fully updated each year. In spring, there is a detailed requirements analysis, composed by the DLCL on the bases of in-depth discussions with the community, which gives an extensive overview on the current and the aspired state of each data life cycle and summarizes the steps to be taken to achieve the latter. During summer, the project schedule for the upcoming 18 months is reviewed and adjusted to the needs described in the detailed requirement analysis and to the latest technical state of the art.
Starting the second year, there is a community forum every year, where communities exchange their experiences and articulate their requirements and wishes. Each spring, there is an all-hands meeting; in fall, there is an annual international LSDMA symposium, bringing together big data experts from all over the globe and LSDMA collaborators.
V. C ONCLUSIONS Recognizing data as source for new knowledge, the deployment of five Data Life Cycle Labs within the LSDMA project of the German Helmholtz Association is well underway. By combining the competence of several research centers, universities and communities in a novel superstructure, an accepted and efficient research and development platform for data management of the data life cycle is successfully established. The outlook towards a generic infrastructure in which all Helmholtz centers participate ensures a long term support for many scientific communities and gives the project a leading role in Germany.
The core building block is the Data Life Cycle Lab (DLCL) in which data experts closely work together on joint R&D topics with community scientists. Together they address the iterative optimization of community-specific data life cycles, the creation of services for simple access and use of data and local, national and international infrastructures for data storage, data processing and data archival as well as for international collaboration. The DLCLs are complemented with by a Data Services Integration Team (DSIT), where a specialized team of data experts work on generic, multi-community data management and analysis services and – together with the DLCLs – the integration of these services in the data life cycle of the scientific communities. The developed generic components and tools can be reused and applied for a growing range of scientific disciplines, allowing faster results and compliance with requirements for publishing results of scientific activities.
Already in its first year of existence, the project has gained considerable momentum in the scientific community and new
communities are interested to establish DLCLs for their specific needs.
With its concept LSDMA facilitates scientific communities through advanced data exploration and preservation services with faster data extraction from original, derived and archived data, thereby immediately gaining scientific insight and enabling new societal knowledge in the future.
ACKNOWLEDGMENT The authors wish to thank all people and institutions involved in defining and setting up the LSDMA project as well as the German Helmholtz Association for providing the funding.
R EFERENCES [1] J. Gray. eScience Talk at NRC-CSTB meeting Mountain View CA, 11 January 2007, http://research.microsoft.com/en- us/um/people/gray/talks/NRC-CSTB_eScience.ppt [2] Grid Computing Centre Karlsruhe, http://www.gridka.de/ [3] Worldwide LHC Computing Grid, http://wlcg.web.cern.ch/ [4] A. García, S. Bourov, A. Hammad, J. van Wezel, B. Neumair, A. Streit, V. Hartmann, T. Jejkal, P. Neuberger, R. Stotzka. The Large Scale Data Facility: Data Intensive Computing for scientific experiments. Proceedings of The 12th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC-11/IPDPS- 11), IEEE Computer Society, 1467-1474 (2011), http://dx.doi.org/10.1109/IPDPS.2011. [5] N. Attig, R. Esser, P. Gibbon. Simulation Laboratories: An Innovative Community-Oriented Research and Support Structure. In: CGW’07, ACC CYFRON ET AGH, 2008, ISBN 978-83-915141-9-1, pages 1- [6] T. Hey, S. Tansley, K. Tolle. The Fourth Paradigm: Data-Intensive Scientific Discovery. Microsoft Research, ISBN 978-0982544204, http://research.microsoft.com/en-us/collaboration/fourthparadigm/ [7] M. Riese, F. Friedl-Vallon, R. Spang, P. Preusse, C. Schiller, L. Hoffmann, P. Konopka, H. Oelhaf, T. von Clarmann, M. Höpfner. GLObal limb Radiance Imager for the Atmosphere (GLORIA): Scientific objectives. Adv. Space Res., 36:989–995, 2005. [8] M. Endemann. MIPAS instrument concept and performance. In Proceedings of ESAMS ’99 – European Symposium on Atmospheric Measurements from Space, pages 29–43, 1999. [9] ESA Envisat, http://www.esa.int/Envisat [10] E. Guilyardi, V. Balaji, S. Callaghan, C. DeLuca, G. Devine, S. Denvil, R. Ford, C. Pascoe, M. Lautenschlager, B. Lawrence, L. Steenman- Clark, S. Valcke. The CMIP5 model and simulation documentation: a new standard for climate modelling metadata. CLIVAR Exchanges Special Issue No. 56, Vol. 16, No.2, May 2011 [11] F. Giorgi, C. Jones, G.R. Asrar. Addressing climate information needs at the regional level: The CORDEX framework. World Meteorological Organization Bulletin 58(3), 175–183, 2009 [12] M. Lautenschlager. Institutionalisierte „Data Curation Services“, S. 149-
- Aus: Handbuch Forschungsdatenmanagement, Herausgegeben von Stephan Büttner, Hans-Christoph Hobohm, Lars Müller, BOCK + HERCHEN Verlag Bad Honnef 2011 [13] M. Greenwald, D. P. Schissel, J. R. Burruss, T. Fredian, J. Lister, J. Stillerman. Visions for Data Management and Remote Collaboration on ITER. Proceddings of the 10th International Conference on Accelerator and Large Experimental Physics Control Systems, Geneva, Switzerland (2005) [14] International Thermonuclear Experimental Reactor, http://www.iter.org/ [15] Massoud Amin, S. and Wollenberg, B.F., Toward a smart grid: power delivery for the 21st century. Power and Energy Magazine 3/5, 34-41, IEEE [16] M. Mültin, F. Allerding, H. Schmeck. Integration of electric vehicles in smart Homes - An ICT-based solution for V2G scenarios. IEEE PES Innovative Smart Grid Technologies, Washington DC USA Jan 2012.
[17] C. Daniel, J. O. Besenhard. Handbook of Battery Materials, Second Edition. 2011 Wiley-VCH Verlag GmbH & Co. KGaA [18] http://www.fz-juelich.de/inm/inm- 1/EN/Forschung/_docs/BrainMapping/BrainMapping_node.html [19] K. Zilles, K. Amunts. Centenary of Brodmann’s map – conception and fate. Nature Reviews Neuroscience 11(2): 139-145. [20] The ANKA synchrotron facility, http://ankaweb.fzk.de/ [21] R. Stotzka, W. Mexner, T. dos Santos Rolo, H. Pasic, J. van Wezel, V. Hartmann, T. Jejkal, A. Garcia, D. Haas, A. Streit. Large Scale Data Facility for Data Intensive Synchrotron Beamlines. 13th International Conference on Accelerator and Large Experimental Physics Control Systems, 2011 [22] P. Lemmer, M. Gunkel, D. Baddeley, R. Kaufmann, A. Urich Y. Weiland, J. Reymann, P. Müller, M. Hausmann, C. Cremer. SPDM – light microscopy with single molecule resolution at the nanoscale. Appl. Phys. B 93: 1-12, 2008Müller, P.; Schmitt, E.; Jacob, A.; Hoheisel, J.; Kaufmann, R.; Cremer, C. & Hausmann, M. (2010), 'COMBO-FISH enables high precision localization microscopy as a prerequisite for nanostructure analysis of genome loci', Int J. Molec. Sci. 11, 4094-4105. [23] N. Becherer, H. Jödicke, G. Schlosser, J. Hesser, F. Zeilfelder, R. Männer. On Soft Clipping of Zernike Moments for Deblurring and Enhancement of Optical Point Spread Functions. In 'Image Processing: Algorithms and Systems, Neural Networks, and Machine Learning, Proceedings of the SPIE', pp. 73-83. [24] CellNetworks, http://www.cellnetworks.uni-hd.de/ [25] K. Schultheiss, J. Zach, B. Gamm, M. Dries, N. Frindt, R.R. Schröder, D. Gerthsen. New electrostatic phase plate for phase-contrast transmission electron microscopy and its application for wave-function reconstruction. Microsc Microanal, 16, pages 785- [26] A. Kobitski, J. C. Otte, B. Schäfer, U. Strähle, U. Nienhaus. Development of digital scanned laser light sheet microscopy with high spatial and temporal resolution. Accepted Presentation at the "Annual Meeting of the German Biophysical Society", 23-26 September 2012, Göttingen, Germany
[27] GSI Helmholtzzentrum für Schwerionenforschung GmbH, http://www.gsi.de/en/ [28] FAIR - Baseline Technical Report, Volume 1-6, September 2006, ISBN: 3-9811298-0-6, for a pdf version see: http://www.fair-center.eu/fair- users/publications/fair-publications.html [29] CRISP http://www.crisp-fp7.eu/ [30] D. Protopopescu, K. Schwarz, PandaGrid - A Tool for Physics (PS36-2-
- Journal of Physics: Conference Series {\bf 331} (2011) 072028) [31] ALiEn, http://alien2.cern.ch/ [32] FairRoot, http://fairroot.gsi.de/ [33] HEP-SPEC’06 Benchmark, http://w3.hepix.org/benchmarks/doku.php [34] dCache, http://www.dcache.org/ [35] XRootD, http://xrootd.slac.stanford.edu/ [36] gLite File Transfer Services (FTS), https://www.gridpp.ac.uk/wiki/GLite_File_Transfer_Service [37] M. Sutter, V. Hartmann, M. Götter, J. van Wezel, A. Trunov, T. Jejkal, R. Stotzka. File Systems and Access Technologies for the Large Scale Data Facility. In: Remote Instrumentation for eScience and Related Aspects, Springer, 2012, ISBN 978-1-4614-0508-5, pages 239-256, http://dx.doi.org/10.1007/978-1-4614-0508-5_ [38] M. Hedges, T. Blanke, A. Hasan. Rule-based curation and preservation of data: A data grid approach using iRODS. Future Generation Computer Systems, Vol. 25 Iss. 4, Elsevier, 2008, pages 446-452, http://dx.doi.org/10.1016/j.future.2008.10. [39] T. White. Hadoop: The Definitive Guide. O’Reilly, 2010, ISBN 978- 1449389734 [40] EUDAT, http://www.eudat.eu/ [41] Concept for improving Supercomputing and Big Data in the German Helmholtz Association, in German, not yet publicly available [42] e-IRG-HLG (High Level Expert Group on Scientific Data). Riding the Wave: How Europe can gain from the rising tide of scientific data. European Commission, 2010, http://cordis.europa.eu/fp7/ict/e- infrastructure/docs/hlg-sdi-report.pdf
