Academia and vendors discuss merging of Grid, Cluster and Supercomputing at ISC2002 TutorialHeidelberg 21 jun 2002 The ISC2002 Tutorial on Clusters and Grids was the best attended tutorial ever, as stated by Professor Dr. Hans Meuer on June 19th. The presentations of academic researchers and vendors attracted no less than 98 participants which was mainly due to the interesting topic and the well known speakers. The general perspective in the field of supercomputing is that the two fields of cluster and grid computing are merging together.
The tutorial brought together two parties who are involved with HPC computing from a different angle: the major vendors with their products and academia focusing on research and prototypes. In the morning, the speakers addressed the hardware and low level systems whereas in the afternoon, the main problems in resource management were dealt with.
Dr. Alexander Reinefeld from ZIB and Humboldt University in Berlin introduced the topic of Clusters and Grids and offered a short overview of past, current and future developments. HP speaker Frank Baetke held a talk on next generation nodes as building blocks for upcoming Clusters and Grids. SGI's visualisation expertise in relation to the Grid was highlighted by Rüdiger Wolff.
IBM representative Jamshed Mirza explained what the eLiza project has to offer to Grid computing. In turn, Martin Walker described how Compaq is powering the Grid. The morning sessions ended with a view on what is happening in the Kirchhoff Institute for Physics at the University of Heidelberg. Volker Lindenstruth demonstrated to the audience how reliable HPC Clusters can be built with over thousands of nodes. The possibility to build reliable big clusters was one of the most important topics of conversation during the tutorial.
Another major discussion which regularly occurred across the day involved the total cost of ownership in the sense of the manageability of Clusters. In this regard, the audience had particular concerns with regard to the manageability of large Clusters. In fact, the tutorial showed a give and take between vendors and academia and provoked a lively discussion between these two groups.
The afternoon sessions started with a presentation from Ulrich Brüning on ATOLL, a network card which has been developed, not by a vendor but at the University of Mannheim. Wolfgang Hafemann from the IWR Heidelberg highlighted the selfmade commodity Cluster consisting of 512 nodes which has been developed at Chemnitz. For NEC, Guy Lonsdale outlined the activities and perspectives for Cluster and Grid computing aimed at by this company.
The subject of resource management was specifically treated by Wolfgang Gentzsch when talking about the Sun Grid Engine and by Songnian Zhou from Platform Computing. The last talk was held by software vendor Pallas to present the Unicore software to the audience.
During the final plenary discussion, issues such as power and heat, flow space and buildings for large-end clusters were highlighted. Once again, the growing together of Grid, Cluster and Supercomputing was emphasised. The audience received a clear idea of what vendors have to offer in terms of Cluster and Grid computing but was pleasantly surprised to learn that academic researchers are very active in the field and currently come up with a host of exciting ideas and practical developments.
Leslie Versweyveld |
 |