July 2008 (Vol. 9, No. 7) p. 2 1541-4922/08/$26.00 © 2008 IEEE Published by the IEEE Computer Society Distributed Computing Education, Part 2: International Summer Schools
In a previous installment of this department (http://dsonline.computer.org/portal/pages/dsonline/2008/06/o6002edu.html), we discussed several challenges facing educators in distributed computing. The distributed computing community has begun to address these problems through the International Summer Schools in Grid Computing (ISSGC) series (www.iceage-eu.org/issgc08/index.cfm), started in 2003 and now in its sixth year. The schools recently gained support from the EU FP6 ICEAGE Project (www.iceage-eu.org). These schools bring together leading experts and practitioners in the Grid computing field with selected international students for two weeks of intensive study. The selection procedure gives special consideration to students who describe how they will use their experience at the summer school to disseminate expertise locally and within their field of study. The aim is for students to pass their knowledge and enthusiasm on to others. The age distribution of attendees suggests that the majority of students are at the stage of early postdoctoral research or initial academic teaching appointments. Student characteristics The number of students at the schools has varied over the years, with a peak of approximately 80 students. However, in recent years we have limited attendance to provide an optimal teaching environment. Almost every year, the school is over-subscribed by twice the number of students we can actually accept; this is determined by the ability to provide on-site facilities. Students come from all over the world, but their geographical distribution shows a bias toward Europe, where all of the summer schools to date have been held. Obviously, the cost of attendance for students from Europe is much lower than for those from other areas. Table 1 shows how many students from various geopolitical regions attended ISSGC 07, held in Sweden. It shows the figures for the entire registration process. The "Started" column lists the students who began the process. "Submitted" provides numbers of students who submitted a registration form with referee letters. "Offers" shows the number of offers made to students to attend, and "Attended" lists the number of students who actually came to the summer school. Table 1. Participation by geographic location. Table 2. Participation by gender. The clear gender imbalance detailed in table 2 is a disappointment, of course, but it is typical of the technical community engaged in grids. In 2007, we noted a slight increase in female participants. As table 3 shows, however, the majority of participants come from computer science, physics, and engineering, domains with a well-known male bias. Increased participation from other fields, such as biology or medicine (which is becoming apparent over time), might go some way toward redressing this imbalance in the future. Table 3. Participation by discipline. The curriculum As stated under the learning goals for ISSGC 07, when students have completed the summer school course, they will The summer school curriculum has always aimed for a mix of conceptual and practical teaching (see table 4 ). The pattern that has evolved includes an introductory talk that sets the day's topic in a broad conceptual framework and links the topic to those of the preceding and following days. The rest of the day's teaching tends to include lectures in the morning and practical work in the afternoon. In each summer school, the aim has been to present four to five of the leading technological implementations in the field as specific examples of the topics presented (for instance, specific solutions to the problems of security in a distributed environment). To further encourage a cohesive intellectual structure for the school as a whole, the ISSGC Practical Subcommittee strongly encourages the presenters to include an advanced practical in their topic. This presages a combined group exercise featuring all of the relevant implementations in the second week of the school. Table 4. Program, International Summer Schools in Grid Computing 07. The integrating exercise A unique feature of ISSGC is the use of an integrating exercise as the culmination of the lessons. This exercise demonstrates the features of each implementation the students have encountered in the program, in a setting in which the technologies demonstrate their abilities to support a common task. The central task is based on the common parameter sweep type of problem, which is encountered in many research domains. The students must find 11 "pillars" (see figure 1 ); on top of each pillar is a "plaque" with text. Once found, the first letter of each piece of text forms another word. The students report the text on each pillar (allowing for authentication of the results) and then report the final text. Figure 1. The integrated practical data space. Figure 2 shows the visualization of a pillar using GNUplot. This was the main method for visualization in the 2006 exercise and was still available to students in the 2007 exercise. However, ISSGC 07 also saw the introduction of a new software component that the students could incorporate into their own programs to provide visualization (see figure 3 ). Because the items of data the students are seeking in the exercise are distributed across the entire data space, they must use each of the technologies to find all of the items. Figure 2. Using GNUPlot for visualization. Figure 3. Radar tool for visualizing search results. Figure 4 shows the distribution of the computed (y > 0) and stored (y < 0) areas within the whole data space. Areas not marked as "GT4" or "Condor" outside the computed area were accessible using OGSA-DAI (Open Grid Services Architecture Data Access and Integration), which presented an XML database. Thus, the exercise encompasses both computational and data access aspects of using grid technologies. This gives the students an understanding of the mechanisms used to support both simulations (computationally derived data) and instrumentation (characterized by file- or database-held information). Technology performance may vary between these two domains, and you can't safely extrapolate experience in only one to the other. Figure 4. Computational and database details of the data space. Outcomes of the integrating practical For the practical, the ISSGC Practical Subcommittee chair assigns the students to groups. The students give a final group presentation describing their solutions, but more importantly relating and analyzing the management and functioning of their group. The exercise allows direct comparisons in using different implementations. It also demonstrates the requirements for collaborating in a team to complete a task too large for any individual to manage, as well as the techniques required to manage such teams. Figure 5 shows the rates at which the groups discovered pillars for each of the technologies, and figure 6 presents the overall time it took each group to find the pillars across technologies (or regardless of technologies). In general, groups started with Condor and found this the easiest technology to grasp early on. The second technology approached by groups was either Globus or OMII-UK (Open Middleware Infrastructure Institute-UK ) software. Groups generally didn't attempt to use gLite until they had been successful with other technologies. However, the groups that did use it noted that the remote resources weren't as saturated as the other technologies' local resources. Figure 5. Team working and exercise evaluation. Figure 6. Time of groups finding pillars. Student response and evaluations Figures 7 and 8 present results from the student evaluations, in which items are listed by obtained scores. Each horizontal bar shows the average evaluation for an item, on a scale from 1 (poor) to 6 (excellent). Each bar segment's length indicates the proportion of evaluations for that item at each point on the scale. The first diagram shows how strongly students felt that the school had achieved its goals. The second shows an evaluation of the general aspects of the school. Out of 62 students, 52 responded to evaluate the first week of the summer school, which represents a response rate of approximately 84 percent. We received 46 responses regarding the second week and overall feedback. The lower rate for the second questionnaire reflects the fact that some students left early on the last day. Even for an enclosed event, this is a remarkably good response rate. It shows student engagement with the school and the efforts its organizers made. Figure 7. Student evaluation of how well ISSGC 07 achieved its goals. Figure 8. Student evaluation of general aspects of ISSGC 07. The overall assessment of the school was approximately 5.29 out of 6, up from 4.68 for ISSGC 06 and 3.67 for ISSGC 05, when the measurement of this parameter started. A continuing rise of this score in this type of survey is remarkable and indicates a major improvement in the students' perceived quality of the school. Conclusion By imparting valuable skills and knowledge in grid computing, ISSGC prepares students to share their newly acquired expertise and to promote e-infrastructure technologies in their home communities. Students unable to attend ISSGC 07 had an opportunity to participate in a similar experience in February 2008, when the first International Winter School on Grid Computing (IWSGC 08) was held online; the course got around barriers such as cost, location, and limited spaces that might have prevented students from attending the summer schools. The next article in this series will provide details of the winter school experience. David Fergusson is deputy director of training, outreach, and education at the National e-Science Centre, Edinburgh. Contact him at dfmac@nesc.ac.uk. Richard Hopkins retired, was a trainer at the National e-Science Centre, Edinburgh. Contact him care of dfmac@nesc.ac.uk or evmeer@nesc.ac.uk. Diego Romano is a researcher in computational and computer sciences at the University of Naples Federico II. Contact him at diego.romano@dma.unina.it. Elizabeth Vander Meer is the education and training policy officer at the National e-Science Centre, Edinburgh. Contact her at evmeer@nesc.ac.uk. Malcolm Atkinson is director of the e-Science Institute and e-science envoy at the National e-Science Centre, Edinburgh. Contact him at mpa@nesc.ac.uk.
| ||||||||||||||||||||||||||||||||||||||||||||||||||
DOWNLOAD PDF 
be familiar with Grid environments' fundamental components, such as authentication, authorization, resource discovery, and resource access; 
