EGEE: Building a Pan-European Grid Training Organisation Prof. Dr. Malcolm Atkinson Dr. David Fergusson National e-Science Centre 15 South College Street Edinburgh EH8 9AA United Kingdom {mpa,dfmac}@nesc.ac.uk Dr. Rüdiger Berlich* Marcus Hardt, Dr. Marcel Kunze Institute for Scientific Computing Hermann-von-Helmholtz-Platz 1 76344 Eggenstein-Leopoldshafen Germany {berlich, hardt, kunze}@iwr.fzk.de Abstract 2 As the largest EU Grid project, it is one of the Enabling Grids for E-SciencE initiatives foremost goals to establish a production quality Grid infrastructure. Amongst EGEE’s diverse duties the creation of a panEuropean Grid training organisation plays a prominent role. Close to the end of the first project phase, this paper describes the training organisation in the context of EGEE’s other duties, discusses the experiences made in building it, lists results and lessons learned over the course of almost two years of training provision. It is the purpose of this paper to allow upcoming Grid initiatives to leverage the experiences made in EGEE training.1 EGEE comprises research activities (“JRA”), service offers (“SA”) available to existing users, and networking activities2 (“NA”). The NA3 (“user training and induction”) group forms a core part of what is often referred to as the “Virtuous Cycle” (see figure 1). In this concept, a new scientific community is established by dissemination and outreach activities (NA2). Specific requirements of this group’s applications find their way into the EGEE middleware (JRA1). An established user community is formed by training the new users in the deployed applications and middleware (NA3). This community can then in turn help to recruit new communities (NA2). Training is of particular importance in this concept due to its ability to actively form a “critical mass” of users willing and capable of acting in a Grid environment. As Grid Computing still is in the “early adopter” stage, creating a strong user-community early on has a direct impact on the adoption rate of this techology in the “late majority” stage. 1 Introduction: The EGEE Initiative Grid Computing has, over the past few years, matured sufficiently to make it a viable solution for realworld problems. However, there are many different toolkits today that allow to build a Grid environment. And while this wealth of different solutions may offer techniques applicable to the widest possible range of computational problems, their very availability directly contradicts the inherent promise of the ”World Wide Grid” to offer a compatible and standardised computing environment. Similarly, a multitude of different Grid organisations caters only for the requirements of particular projects instead of the whole of the Grid community. EGEE (”Enabling Grids for e-SciencE”) (1 ) is the European Union’s largest Grid initiative. While the European origin implicates a geographically limited scope, EGEE has today outgrown its European origins, with many partners being based in nonEuropean countries, such as Israel, Russia or the United States of America. It is the goal of the EGEE project to create a production quality Grid infrastructure for user groups as diverse as High Energy Physics and Bioinformatics. To achieve this goal, a “best-of-breed” middleware – gLite – is complemented by a services- and supportinfrastructure. Among other duties, EGEE offers a Global Grid User Support, runs Regional Operation Centres and organises training events for interested parties. EGEE thus tries to overcome the problems posed by the various technical solutions and diverse communities. 1 c Copyright 2006, Australian Computer Society, Inc. This paper appeared at Fourth Australasian Workshop on Grid Computing and e-Research (AusGrid 2006). Conferences in Research and Practice in Information Technology, Vol. 54. Rajkumar Buyya and Tianchi Ma, Ed. Reproduction for academic, not-for profit purposes permitted provided this text is included. Training and the “Virtuous Cycle” 3 Organisation, Funding and Infrastructure Training is carried out by 22 of EGEE’s 70 international partners under the leadership of the University of Edinburgh (UEDIN) in the United Kingdom. A total of 22,19 FTEs (“Full Time Equivalent”) equivalent to approximately 1 FTE/partner are in part funded by 4% of EGEE’s 32 Million Euro two-year budget. The 1.3 Million Euro budget is in many cases complemented by monetary contributions by the local partners. UEDIN as well as partners contribute training material, such as PowerPoint or OpenOffice slides to a common repository available to all partners. Existing course material thus forms the building blocks for new courses, and undergoes a constant re-engineering process. The goal is to achieve an evolutionary improvement of material, as trainers rework presentation slides and incorporate changes or fix mistakes. Where necessary, material is translated into local languages. Additional support is available to partners through pre-fabricated registration pages on the NeSC web server. This offer further reduces the amount of work going into the preparation of a training event, so partners can use this time for more important work. As it is impossible for partners to maintain expertise in all possible Grid training areas, trainers of other partner organisations can participate in the more important events. This is particularly true for 2 In the sense of “social networking”, not “data transport” Figure 1: The EGEE virtuous cycle – A new scientific community is established by dissemination and outreach activities. Specific requirements of this group’s applications find their way into the EGEE middleware. An established user community is formed by training the new users in the deployed applications and middleware. This community can then in turn help to recruit new communities. summer schools, which tend to be international by design. Partners can exchange experiences through a mailing list and during regular meetings (usually held in conjunction with the half-yearly EGEE project conferences). Events are monitored and reported to the activity leader on a case-by-case basis. As is common in EU projects, quarterly reports from all partners summarise their activities. These reports then find their way into “deliverables”, documents that allow the EU to monitor, whether a given set of mile-stones has been reached. Deliverables are reviewed internally by another EGEE workpackage before being officially submitted to the EU. In addition to the EU audit, the NA3 group can choose to perform additional quality checks through an external review, and has done so in the past. EGEE and thus the NA3 group is organised in a matrix type organisation, as is necessary in an environment, where independent research institutions from all over Europe strive at achieving a common goal. It is thus necessary for all partners to agree on a common goal, as only very little disciplinary power lies in the hand of the activity manager. Apart from using local channels, advertising for training events can be done through the EGEE site. Both the NA2- and NA3-groups advertise events on their web pages. News- and press-releases can be used for the most importan events, such as Grid computing schools. The technical backbone of EGEE training is formed by the t-Infrastructure. Courses require a labenvironment, in which participants can experiment with the new techniques. While this environment was initially formed solely from local or federation-wide ressources, NA3 strives at the creation of an activitywide t-infrastructure available across all federations. This infrastructure may then be complemented by local computing environments, such as laptops or a computing lab needed to access the Grid. The most widely used environment in EGEE training at the moment is the Gilda testbed (3 ), which is run as a cooperation of the NA3 (“User Training and Induction”) and NA4 (“Application Identification and Support”) groups. Gilda stands for “Grid Infn Laboratory for Dissemination Activities”. It consists of a series of sites, not limited to INFN and Italy, a Ressource Broker, Information Index, Data Managers and Monitoring Tools, Computing Elements at each site and Storage Elements. Currently the testbed runs both the LCG-2 and the gLite middlewares3 . Through the Genius interface, selftraining is offered even to users not participating in an organised training. INFN Catania is the main organiser of Gilda. 4 Training Clientele EGEE’s primary user groups come from the high energy physics and bioinformatics areas. With their already existing need for Grid techniques, establishing the Grid as part of their core computing infrastructure is at the top of their agenda. Projects like the upcoming Large Hadron Collider (“LHC”) at CERN/Geneva will not be able to satisfactorily perform distributed analysis of the estimated 40 petabyte of data accumulated each year4 without an already established Grid infrastructure. With Monte Carlo production for LHC and data challenges already running, preparations for the LHC Computing Grid are well underway. LHC members need to already have developed an understanding of Grid technologies to do their work. A similar situation can be found in Bioinformatics. Not surprisingly, therefore the main clientele for Grid trainings in EGEE so far came from these two environments. Typical participants in Grid trainings include advanced graduate students, PhD students and researchers from universities and independent research institutions. New EGEE communities include Astronomy5 , Chemistry, Earth Observation and Geophysics. As these groups can still perform their research without the use of Grid technologies, participants from these groups will join more likely due to personal interest than due to an immediate need. There is a large variety of different Grid types. Industrial usage of Grid technologies is thus not limited to the “global batch submission type” common in EGEE. Only a limited amount of participants in EGEE trainings thus come from an industry environment. Likewise, there seems to be close to no commercial participation in running Grid courses. 3 4 5 LCG-2 and gLite are the two middlewares in use EGEE with about 6000 physicists worldwide trying to access it ... e.g. the MAGIC project While the NA3 group has given trainings for members of the EGEE project (such as induction courses for the NA2 group or installation courses for the technical EGEE community), the majority of participants in EGEE training courses held by the NA3 group come from organisations unrelated to EGEE. This tendency is likely to be upheld for the second project phase of EGEE scheduled for April 2006. 5 Course Types As can be expected in an early-adopter environment, EGEE’s Grid courses can be broadly grouped into one of the following categories: localised material would certainly be the ideal situation, lack of ressources and a fast-paced technical environment does not allow for continuous translation of new material. It is the intention of the NA3 group to continuously improve the quality of its trainings. Anonymous, standardised feedback forms are thus handed out to participants after each training, and the numerical evaluation of these forms is then handed back to the lead organisation that is responsible for quality assurance. Usually the following parameters (rated in six steps from poor to excellent) are measured: • “Overall evaluation”: How did a participant rate the entire event ? • Induction courses – Their purpose is to give participants an initial understanding of Grid technologies. Part of these courses could be an introduction to the command line interface of a Grid middleware or the usage of a Grid portal such as Genius. These courses tend to also show a large overlap with the activities of the NA2 group, as induction courses for a technical audience could as well be perceived as dissemination. This type of course can be a useful tool for finding new types of user communities. A general technical interest in Grid technologies can be found in many areas, not the least due to the fact that “The Grid” has become a buzz word. A deaper understanding of Grid computing can however reveal new, hitherto unexpected use cases. • Advanced courses – These courses teach advanced topics such as the installation or maintenance of a Grid middleware. They are targeted at people wishing to administer a Grid environment rather than at (potential) Grid users. • Application Developer Courses – These courses teach how to write Grid applications or how to port an existing application to the Grid. In EGEE, an introduction to the APIs of EGEE’s Grid middlewares is usually complemented by an introduction to the field of Webservices. • Technical Retreats and Workshops – These are not courses in the traditional sense, but provide a means for likeminded, technically astute people to work on joint solutions for pressing issues. • Train the trainer – Existing trainers can learn about new technologies in this type of course and new trainers can learn about EGEE’s technologies and the established tools and procedures made available by the EGEE training group. 6 Training Preparation and Training Duties Training material prepared by individual partners should follow the general design guidelines prepared by the NA2 (“Dissemination and Outreach”) group. PowerPoint templates exist that can be used for the preparation. Existing material from the training repository can be used as building blocks for a particular course, although no general rule exists that enforces this. Changes to the material should however be fed back into the repository. Partners are under no obligation to localise the content of the curriculum. It is a common procedure to use the English version of the training material while speaking in the native language of the country where the training is held. While the EU is a multilingual environment and the common availability of • “Achievement of goal”: This could be similar to “Learning how to install the gLite middleware” • “Presentations”: How good or bad were the talks and presentations ? • “Event material”: Can the handouts be improved ? • “Lab exercises”: Were the hands-on components of the course accepted by the audience ? • “Facilities”: Was the training room appropriate ? • “Advertising and Registration”: How easy was it to find out about this event and to register for it ? Additional questions like “What did you like most/least about this event” and comment sections for each topic provide fine-grained, quantitative feedback about the audience’s opinion. Analysis of these results can be done with standard statistical methods. Figure 5 shows a typical evaluation of a course held in conjunction with GridKa School 2004 at Forschungszentrum Karlsruhe (Germany). Getting all attendees of a course to fill out the forms can be a challenge. An approach common in industry, to hand out little presents (USB sticks, ...) in return for the evaluation forms, is not possible in courses that are offered free of charge. In some cases it is possible to swap a certificate of attendance for an evaluation forms. In other cases one has to rely on good will. It does help to ask each participant in person to fill out the form, when walking through the lines to check for questions or to help with problems. It also helps to hand out feedback forms early in the course. Due to the importance of objective feedback to the constant improvement of trainings provided, the area of feedback-analysis should not be neglected. 7 Results from the First Project Phase With the final project conference being held in Pisa/Italy by the end of October 2005, EGEE nears the end of the first project phase6 . Despite the fact that the EGEE training team had to start from scratch and, in some cases, trainers coming from an academic environment have had little to no prior exposure to running trainings, the EGEE NA3 group has had a number of remarkable successes. Indeed, many of NA3’s targets have already been exceeded, and both internal and external reviews acknowledge the success of the group. 6 It is likely that EGEE-II will take over from EGEE in March 2006 Course Type Induction Developer Advanced Workshops Participant Days 1541 436 1522 741 1-year Target 1000 800 250 360 % of 1-Year Target 154 54 609 206 Table 1: Participant days at the end of project month 15 in comparison with the one-year target Figure 2: Accumulated number of students as a function of time until May 2005. Figure 4: Photograph of an EGEE introductory course for the LCG-2 middelware, held in conjunction with GridKa School 2004 at Forschungszentrum Karlsruhe. The EGEE Training Progress Report (2 June 2005) lists many of these results. During the first 15 months7 of the project, a total of 2019 people have attended 131 single training events equivalent to 4864 participant days. Table 1 lists these figures in comparison with the targets for the first year, separately for different training categories. With the exception of Application Developer courses, thanks to the enthusiasm and hard work of all members of EGEE, all course types exhibit a significantly higher amount of participant days than the initial targets. Reasons for the Developer courses’ lower outcome include the higher complexity and the higher trainer-to-participant ratio required to run such courses. Still, demand for Developer courses already exceeds supply. As is shown in figure 2, the overall number of participants in the courses is constantly increasing, albeit with a declining slope. Remarkably, as is shown in figure 3, the overall rating of the EGEE NA3 courses is constantly increasing as well. In the diagramme, “6” stands for “excellent”, “1” for “poor”. 8 Figure 3: The participant’s ratings of EGEE’s courses is increasing over time, indicating an increase in quality. This tendency would not be visible without constant monitoring of training quality. Section 6 lists some of the criteria checked in the evaluation forms filled out by participants. The plot shows the overall rating attributed by users to the courses. A rating of 1 corresponds to “poor”, 6 means “excellent”. Experiences made in the DE/CH federation Within the German/Swiss (“DE/CH”) EGEE federation, training is handled by Forschungszentrum Karlsruhe (“FZK”). As is common for many other federations, DE/CH has attributed 1 FTE to NA3 of which 50% is funded by the EU, the other half by FZK. Likewise, NA2 (“dissemination / outreach”), which is also handled by FZK, is equipped with 1 FTE. For practical reasons, dissemination/outreach and training are treated as a unity at FZK. One person has the main responsibility for both workpackages, and work contributions up to 0.5 FTE in size account for the remaining 1 FTE. 7 At the time of writing this document, only the results up to August 2005 are known Figure 5: Typical evaluation of a course held in conjunction with GridKa School 2004 at Forschungszentrum Karlsruhe. Due to these limited resources, a decision had to be taken early in the project to initially attribute more time to training than to dissemination and outreach. The idea was that by training people to use Grid technologies early, a strong basis of multipliers could be generated that dissemination/outreach could build on during the second year of the first project phase. Also, in an early-adopter environment, training and dissemination share many characteristics. Under the assumption that the vast majority of participants in early Grid trainings would understand English well enough, and in order to free man-power for giving the actual trainings, a decision was taken not to translate training material into German language. Trainings were nevertheless given in German language (but with English slides, hand-outs, etc.) in those cases, where only participants were present that spoke German well enough to follow the course. FZK is home to the largest German scientific computing cluster (“GridKa”), and active in a number of other EGEE workpackages, as well as other Grid projcets (e.g. D-Grid, CampusGrid, and until recently also CrossGrid). GridKa itself is a Tier-1 centre in the LHC Computing Grid. While the training activity is funded relatively low in Germany when taking into account that Switzerland and Germany combined cover about a quarter of the EU population, the FZK NA3 group can benefit from expertise and to some extent also man-power from other FZK activities. The infrastructure available at FZK – one of the largest independent research centres in Germany – in additon allows to offload some tasks, such as large print runs or translation work. It should however be noted that letting people alien to Grid computing do translations of highly complex technical texts can result in an additional work load due to the need for quality assurance, and thus has its limits. Using these services has however proven to be very useful for the more general dissemination material. Due to its close involvement in the CERN Alice Experiment and the fact that early gLite prototypes were largely based on AliEn, there was also a period of close training cooperation with Gesellschaft für Schwerionenforschung (Darmstadt/Germany, “GSI”), another EGEE partner, which contributed its considerable expertise in the AliEn area to a number of trainings. The University of Karlsruhe has helped with trainings, as have EGEE partners (such as DESY or Fraunhofer SCAI) that have hosted trainings organised by FZK. FZK has, at the beginning of the first project phase in April 2004, purchased 10 laptops for training. Together with a 16 node (32 CPU) cluster this forms part of the local t-Infrastructure. The laptops are used as user interfaces or, in some cases (such as webservices courses), also as a computing platform. The cluster is used for installation courses and also serves as a test platform for new technologies. E.g., FZK is involved in the creation of a “Grid in a box”, a virtual Grid environment based on the Xen virtualisation software. In this way, FZK/NA3 also does Grid research inside of EGEE, albeit with a focus on facilitating training. Where technically possible, visiting “customers” for trainings was considered a better approach than asking participants to come to FZK. Only 1-2 people need to travel in this case. The availability of portable training laptops was of big benefit to this concept. A total of 13 courses was given by the FZK group – partially with contributions of man-power and resources by other activities8 – until August 2005 in Germany. As Switzerland joined only late as an unfunded partner in EGEE-I, no trainings could so far been given in this area9 . One of the courses was given in Sydney/Australia, as a cooperation of FZK, GSI, the Australian Grid Forum and the Australian Partnership for Advanced Computing. Three courses were given in cooperation with the National E-Science Centre, one of them in Amsterdam as an internal course for the NA2 group. Internal training of EGEE members in the usage of the LCG-2 and gLite middlewares accounted for two trainings10 . Two installation courses for the LCG-2 middleware were given. As these courses required a very large amount of hardware resources, they were held at FZK itself. Of particular importance is the GridKa school, an annual Grid computing school held at FZK, headed by the FZK NA2/3 group. As it is the only school of its kind in Germany, it regularly attracts a large number of postdocs, advanced undergraduate and graduate students. The 2004 school attracted some 94 participants. The majority of topics of the 4-day school were related to EGEE, including LCG-2 installationand introductory courses. The organisers hope to repeat this success in late September 2005 in GridKa School 2005. Figures 4 and 5 show a photograph of the introductory LCG-2 course at GridKa School 2004 and the corresponding user ratings. Typically, participants in trainings came from a research environment, although an attempt was made in a course held in cooperation with the NA4 group to “market” the EGEE Grid technologies to participants from industry. Finding new user groups can be an uphill battle, but should be considered one of the most prominent duties of the German/Swiss NA3 group in the second project phase of EGEE. An approximate 220 people could be trained in or by the German/Swiss EGEE federation until August 2005. The willingness of other activities and EGEE partners to cooperate and to contribute both manpower and resources to training activities was a key to this success. 9 Experiences made in the UK federation As the lead partner for training within EGEE the National eScience Centre (NeSC) at the University of Edinburgh has had a particular set of tasks and requirements. Not only does it have to provide training for the UK federation of EGEE but also to provide support and leadership within the activity as a whole. It therefore has a larger resource than is common among the other partners. The NeSC training team was formed in April 2004 at the start 8 SA1, NA4, NA3 groups from other federations with the exception of CERN itself, which is not considered part of DE/CH and handles training seperately 10 A few external visitors participated as well 9 of the EGEE project. The first training course aimed at developing an initial set of material and bringing trainers together from across Europe to help stimulate a corporate spirit within the training activity was held in Edinburgh in April 2005. Over the next year the Edinburgh team has been involved in running 17 events in the UK (390 attendees) and a further 16 elsewhere in Europe (355 attendees). These have ranged from introductory events for those with no prior experience of grids to advanced courses, workshops and summer schools catering for those at the cutting edge of grid technologies. To give a flavour of the achievment this represents each course averages about 2 days, generally requires at least 2 trainers and when travel and set-up time is factored in requires about 4 days. From the earlier figures it can be seen that the team averages about 2 courses per month. This means that at least 50% the available time is spent in actually presenting courses. A complicating factor in the UK, from the view point of providing EGEE training, is that while there is a particle physics grid in the UK running EGEE middleware (LCG) called GridPP, the UK National Grid Service (NGS) has not yet fully adopted the EGEE middleware stack. Thus, to a degree, UK grid training material has to be localised to be relevant to those working in a UK context. The University of Edinburgh is also tasked with providing a web site (http://www.egee.nesc.ac.uk/) for the EGEE training activity. This provides a central source for advertising training events, providing registration support services, quality assurance materials and monitoring. The web site also provides access to the EGEE training materials archive. This is an open source/access repository for all training material created in the EGEE project (http://www.egee.nesc.ac.uk/trgmat/index.html). This now contains in excess of 1000 presentations and greater than 2000 files. The repository is now being used as the basis for developing a more comprehensive eLearning framework for the EGEE project which will allow the training activity to deal effciently and effectively with the challenges of scaling with the rapid increase in the size of EGEE, both geographically and in terms of subject domains. 10 Conclusion and Lessons Learned Both the user feedback and the various performance reviews indicate that the work done by the NA3 group is successful. Thus there is no need for radical changes in the design and setup of the group structure in the second project phase. A few moot points remain, however. • The NA3 group had to rely on investments of its partners for the t-infratructure, as it was not assigned dedicated funds for building this environment. A more focussed approach would be possible, if NA3 would posses funds for buying computer hardware. • Some training types, such as the Application Developer courses, suffer from a lack of training personnel. It must be stressed that this is due to principal reasons and not due to a lack of enthusiasm or dedication on the side of the NA3 group. • On a related issue, the Grid environment represents a very fast moving target. With an overall of 22 FTEs working in Grid training inside EGEE during the first project phase and thanks to the high demand for training courses, is can be a difficult task to keep the training repository up-to-date. Also, not all training material is fed back into the repository. • The relatively small number of FTEs assigned to NA3 means that working in the NA3 group can result in very high pressure. It must be stressed that this pressure has not resulted in a lower number of courses given by NA3. • Finding a fair share of FTEs to be assigned to partners can be difficult. Apart from the size of the “training market” in a particular federation, the demand for trainings has to be taken into account. Availability of qualified personnel and synergies have to be taken into account, and strategic considerations are important. In addition the cost per FTE can be very different even for different partners of the same federation. • In an EU project of the size of EGEE, with 70 different partners from all over Europe, political issues will arise and have arisen over the course of the first 15 months. Still, this is the normal environment partners must act in, and it is part of everyday work to deal with these issues. Despite any problems that might have arisen during the first project phase, NA3 has surpassed its targets and can be considered one of the most successful groups of the EGEE project. It should be stressed that support by other EGEE activities was a core factor in this success. It would be desirable for such cooperation to be extended in the second project phase of EGEE. It should be noted that, while there are other organisations in the Grid space that do provide training, an attempt to build a Grid training organisation on this geographic scale, almost entirely based on academic man-power and resources, has not been attempted before. EGEE is also taking over the role of training provider for other Grid-related initiatives. E.g., as LCG-2 has been the first Grid middleware to be used in the EGEE Grid, EGEE has also played a major role in Grid-training provision for LHC-based particle physics. The training schedule (12) also lists Globus trainings and Web services as topics. As early versions of gLite included components of the AliEn middleware, initial trainings in the DE/CH federation of EGEE also covered AliEn in addition to LCG-211 . It is not unlikely that EGEE-II, the successor project to EGEE, will, on a smaller scale, also play a role in UNICORE Grid training (13), due to plans to build bridges to this Grid initiative. As a deployment project, EGEE has to focus as much on interoperability as it does on standards. Training thus has to cover topics beyond the core technical scope of the project. 11 The Authors Dr. Rüdiger Berlich and Marcus Hardt run the training workpackage of the German/Swiss EGEE federation. Dr. Marcel Kunze is the federation representative. Prof. Dr. Malcolm Atkinson is the director of the National e-Science Centre in Edinburgh, Dr. David Fergusson is the EGEE NA3 activity manager. 12 Acknowledgement The authors would like to thank the members of the EGEE federation. All successes presented in this document were the result of the cooperation of many 11 LCG-2 was then the other building block of gLite, and is today still one of its major roots individual people! Likewise, we would like to thank Forschungszentrum Karlsruhe, the German Federal Ministry of Education and Research (BMBF) and the UK National E-Science Initiative for their continuous interest and support. References (1) Members of the EGEE Collaboration EGEE – Enabling Grids for E-sciencE URL http://www.eu-egee.org (2) Members of the EGEE Collaboration EGEE Training Progress Report Month 15 URL https://edms.cern.ch/document/ 533762 (3) INFN Italy The Gilda Testbed URL https://gilda.ct.infn.it (4) Forschungszentrum Karlsruhe URL http://www.fzk.de (5) National e-Science Centre Edinburgh URL http://www.nesc.ac.uk (6) The EGEE NA3 Group Training in EGEE URL http://www.egee.nesc.ac.uk (7) Forschungszentrum Kalrsruhe GridKa School 2005 URL http://www.fzk.de/gks05 (8) Members of the EGEE Collaboration The gLite Middleware URL http://glite.web.cern.ch/glite (9) The EGEE JRA1 group The EGEE Middleware Architecture URL http://edms.cern.ch/document/476451 (10) The D-Grid-Initiative URL http://www.d-grid.de (11) The German Ministry of Education and Research URL http://www.bmbf.de (12) EGEE course schedule URL http://www.egee.nesc.ac.uk/schedreg (13) The UNICORE Grid URL http://www.unicore.org