From: "Bernard F. Schutz" Date: Sun Nov 11, 2001 09:41:47 US/Pacific To: "Barish, Barry" Cc: "Weiss, Rai" , "Lazzarini, Albert" , "Sathyaprakash, B S" , "Allen, Bruce" , "Taylor, Ian" Subject: Grid cooperation with LIGO Reply-To: schutz@aei.mpg.de Dear Barry, Bruce Allen, who is visiting the AEI at the moment, told me that the GriPhyN-based Grid collaboration with the UK is starting up in earnest, and PPARC has offered to put a few postdocs in the US to assist this. This spurs me to write to you about our own Grid projects involving Cardiff/AEI/GEO, which you may not know much about. I think this is a good time to see if we can improve collaboration between the different Grid efforts within the gravitational wave community. The UK offered a lot of Grid money earlier this year, and a chunk went to support UK-US iVDGL collaboration, another chunk went to AstroGrid, and (among other smaller amounts) some went to Cardiff to start up GridOneD. This is initially a collaboration between Sathya's group and the CS people in Cardiff to enhance the functionality of Triana, our Java-based data-analysis environment, to enable it to interface to C code and to turn it into a work-flow program that can supervise and link programs that may be running on a number of remote machines. We see this development as enabling Triana to supervise data analysis and other procedures running with compiled code on any platform, locally or remotely, with a flexible graphical interface. The work is dovetailed to another Grid grant which is centered at the AEI, called GridLab. This is funded by the EU (roughly $4M) and involves a number of European centers. Its purpose is to interface Cactus (our numerical relativity simulation environment) and Triana to the Grid (which means to Globus) and to each other. A group at the computer center ZIB in Berlin is adapting Triana to run on handheld computers. The thrust of this is to provide peer-to-peer Grid computing, including easy distribution of code and management of processing over wide groups. This is targeted at very flexible computing, supporting possibly smaller jobs and smaller platforms than the usual Grid approaches. We see this as useful to GEO. I think it could be attractive within the LSC and, more importantly, it has potential for really making joint global data analysis easier for groups drawn from LIGO/GEO/VIRGO/TAMA.... Triana will become completely public domain by next year, to take advantage of this very significant development effort. The EU grant will put 2 people for 3 years to work on Triana in Cardiff, and a further person at ZIB, not counting people working on common elements of the Cactus/Triana Grid interface. The GridOneD grant will put 2 further postdocs in Cardiff for 3-4 years. Judging from what we see at the Global Grid Forum meetings, Cactus and Triana are likely to be among the first pre-existing applications to have interfaces to a wide range of Grid services. We hope to demonstrate prototypes at the GGF meeting in Scotland next year. For the UK GridOneD grant there is a milestone next year. We have been asked to demonstrate (essentially to Tony Hey) a working prototype of our peer-to-peer model, after which the UK funding could triple or even quadruple. I am writing to you at this point because it seems to me that we should be looking at synergies that could be helpful to both LIGO and GEO. I have been wondering for a while whether there could be a useful way to bring the various Grid initiatives involving gravitational waves closer together. After talking a little with Bruce, it seems to me that there is an obvious way forward. Bruce already has a tier-2 center at UWM, and he spends enough time at the AEI for him to be able to influence the development of GridLab and GridOneD, in particular to help us to steer these projects in such a way that we get a good connection with iVDGL. I suggest that we begin exploring now the possibility of designating UWM as the interface group for Griphyn/iVDGL to GridLab/GridOneD. I doubt that this would divert resources away from their purpose within iVDGL; on the contrary, by helping us to develop an interface to Cactus and Triana it would add value to iVDGL as well as to our projects. I know you have competing demands for resources and never enough people, and that goes for your new UK postdocs too. But if at least one of them could be involved in this interface somehow, then it would make it easier to link your UK collaborators to the Cardiff work, which has so far been totally independent. No doubt Bruce will make this case more strongly to you! I hope the idea of trying to link these various Grid efforts more closely together appeals to you. Best regards Bernard -- Bernard F Schutz Max Planck Institute for Gravitational Physics (Albert Einstein Institute) Am Muehlenberg 1, D-14476 Golm bei Potsdam, Germany Tel: +49 331 567 7220 schutz@aei.mpg.de; http://www.aei.mpg.de Secretary: Fr U Schlichting