ERCIM News No.45 - April 2001 [contents]

The NL Datagrid Initiative

by Kors Bos

---

In the framework of initiatives on the European and world level to realize the Grid concept, the NL Datagrid Initiative proposes to set up an infrastructure in The Netherlands as a test bed for the grid software to be developed by the Dutch participants in the European Datagrid project, and for the software developed in that project as a whole. The national institute for nuclear and high-energy physics NIKHEF plays a key role in this local Grid initiative.

The European Commission has accepted in its IST Programme a proposal to develop a computational and data grid last fall. This follows a number of Grid projects initiated and proposed in North America, in particular the Globus project. A number of European companies and organizations have started the E-Grid Forum (http://www.egrid.org) , paralleling the American organization (http://www.gridforum.org). Recently the two have combined into the Global Grid Forum of which the first meeting took place in Amsterdam on 4-7 March 2001.

Whereas the US Globus project is primarily computational, the EU project focuses on datasets, described in databases where bulk data storage is widely distributed. Areas like particle physics, astronomy, bio-informatics, telecom, and E-commerce increasingly face this situation. The datasets to be processed are not big themselves, but they are many, and very distributed. Three areas serve as a test bed for the software to be developed in the European Datagrid Initiative: particle physics, earth observation, and bio-informatics. The main contractors are: Centre National de la Recherche Scientifique CNRS (France), European Space Agency ESA/ESRIN (Italy), Istituto Nazionale de Fisica Nucleare INFN (Italy), Particle Physics and Astronomy Research Council PPARC (UK), European Organization for Nuclear Research CERN (Geneva, coordinator), and NIKHEF (The Netherlands). Dutch assistant contractors are: Academic Computer Centre in Amsterdam SARA, and the Royal Dutch Meteorological Institute KNMI.

Datagrid testbed sites.

Providing an adequate computing infrastructure is left to the partners in the EU project. In The Netherlands it is estimated that several million guilders are required for this purpose. The work in the EU project is distributed over several Work Packages (WPs). NIKHEF and SARA will develop tools for the Grid middleware WPs Fabric Management and Mass Storage Management, and for the WP Network Services. Not surprisingly, NIKHEF and KNMI participate in the application WPs for High Energy Physics and Earth Observation, respectively.

NIKHEF’s interests in Grid development are derived primarily from the computing needs in the LHC (Large Hadron Collider) era, in view of its involvement in the Atlas, LHCb and Alice experiments. KNMI is, apart from research connected with its operational tasks, a global change research centre. It will use the Datagrid for the exploitation of products from different (satellite) sources and the calculation of, eg, ozone profiles. SARA’s interest is based on its long term strategy, where services will no longer be limited to its own resources. Here complex problems related to data distribution, brokerage, scheduling and resource optimisation, automatic disaster recovery, etc., have to be solved.

The Dutch participants in the European Datagrid project - NIKHEF, SARA, and KNMI - have started the national NL Datagrid initiative to set up an infrastructure in The Netherlands as a test bed for the grid software to be developed by them, and for the software developed in that project as a whole. The infrastructure includes groups of interested people, existing hardware and network facilities. Applications which will serve as a test, are real projects taken from particle physics and earth observation, which are carried out in any case, with or without a grid. SARA plays a special role as a service provider for any scientific enterprise in The Netherlands, and for its data storage expertise. The NL Datagrid initiative envisages a large compute farm at NIKHEF serving all ongoing experiments, as well as local farms at Nijmegen University (D0/Atlas), the Free University in Amsterdam (LHCb) and the University of Utrecht (Alice). The data will be stored at either the tape robot at SARA, at Fermilab, or at CERN. NIKHEF can use resources at SARA such as the new Cray/SGI multiprocessor computer, and data storage in the SARA tape robot. KNMI should similarly be ‘hooked up’ to the high bandwidth infrastructure to profit from facilities elsewhere and make its own resources available to the other partners.

Several local parties have expressed interest in the NL Datagrid initiative, all bringing in their own often very relevant expertise. SURFnet, the National Academic Computing Service provider, considers the Datagrid project as an application candidate to challenge the limit of its network capacity. The SURFnet-5 backbone with a bandwidth of about 20Gb/s is expected to be operational in 2001, thus providing the necessary high bandwidth connection. The Computer Science group of the University of Amsterdam has experience with the Globus tools and contributes to the NL Datagrid initiative through the government sponsored Virtual Laboratory project, in which the Grid will be used for data transfer, processing and storage. The Department of Mathematics and Computer Science of Delft University of Technology is interested in some software aspects of the Datagrid. Five Dutch universities cooperate in the Distributed ASCI Supercomputer project (see elsewhere in this issue). The Datagrid initiative could very well profit from the computer science expertise built up here. CWI has experience in distributed multimedia databases and network management, and is interested in using the Grid for cell biology imaging by virtual reality engines. Finally, the national astronomy foundation ASTRON, hosting the Joint Institute for Very Long Baseline Interferometry in Europe JIVE, is interested in the common use of a broadband fibre network infrastructure (of at least 16Gb/s) for data transport from many radio telescopes in different countries to the data processing centre in Dwingeloo, The Netherlands.

Links:
DataGrid website: http://www.eu-datagrid.org/
NIKHEF: http://www.nikhef.nl/
NL Datagrid Initiative: http://www.datagrid.nl
First Global Grid Forum: http://www.ggf1.nl/

Please contact:
Kors Bos – NIKHEF, Dutch National Institute for Nuclear Physics and High Energy Physics
Tel: +31 20 592 5083
E-mail: kors.bos@nikhef.nl