Project: European Union DataGrid (aka EU DataGrid, EUDG, EDG)
Description: The European Union DataGrid project is one of the largest Grid projects in the world. Its primary goal is to provide the computing infrastructure necessary to deal with the incredible data volume (10s of petabytes per year) that will be generated by projects like the Large Hadron Collider (LHC), which will come online in 2006. The EU DataGrid operates a number of separate Grids -- such as LHC, Earth Observation, and Biomedicine -- that range from a few boxes in a single room to tens of installations across multiple European countries. Once the LHC is online, there will be hundreds of installations and over 2,000 physicists distributed around the globe.
Participants: EU DataGrid Globus (or ANL, ISI) VDT (or UWis), the high energy physics community
Sponsors: European Union, National Science Foundation, Department of Energy
Countries Involved: US, Europe
Tools: The EDG uses the entire gamut of Globus tools: GSI: They have adopted GSI as their underlying security mechanism. They have established their own Certificate Authorities (CAs) and RAs. They are integrating GSI security into their existing applications and of course, all Globus components utilize it. GRAM: Using their own resource broker and the Globus scheduler backends as well as at least one custom back end, GRAM is used to submit jobs across their VO resources. Currently, those jobs are Monte Carlo simulations that will have their output used to tune the filtering parameters on the Hardware filtering of the various ""sensors"" on the collider. Once the LHC comes online, the simulations will continue, but there will also be analysis of the actual collider data as well, with the primary goal the discovery of the Higgs Bosun. MDS: Their resource broker uses MDS to obtain necessary data for making resource allocation decisions. This has been one of the most difficult, but most useful portions of the collaboration. They have exercised MDS at a scale greater than anyone else and have found numerous areas for improvement. GridFTP: GridFTP is the data transport mechanism used to stage data sets to the compute platforms, and to stage the results back to mass storage. This was the earliest adopted piece, as part of the Grid Data Mirroring Package (GDMP). Replication: Also used as part of GDMP, together we quickly discovered that our first implementation did not meet their requirements. We jointly designed the current generation Replica Location Service with the EDG. As of May 2003, EDG just released EDG 2.0 which uses the VDT distribution as a base and includes the customizations necessary for their environment.
Contact: Fabrizio Gagliardi


