The HPC Asia 2009 keynote speech, given by Dr. Peter Arzberger, gives us deep understanding into Dr. Arzberger’s vision for the successful integration of HPC with cross-cultural and international collaboration. He also describes the beginnings of the PRAGMA institute he started several years and currently chairs along with its organizational structure. Dr. Arzberger’s goes on to give us his insight into how to better educate students of science so that they will be better prepared to do research in the global science community.
Part1| Part2 | Part3 | Part4 | Part5 | Part6 | Part7 | Part8 [PDF Download] [Video]
Part 4: PRAGMA Working Groups (cont.)
We’re structured into four working groups. One on Resources, one on Biosciences, one on Telescience, and one on GEO. And what I want to do is walk through an example of each of these to show you how we as a distributed team really have been able to make some progress.
So in terms of resources, one of the things that we’ve done is that we created a test bed in the working group where people can actually run applications and try to understand both how the application responds and also how the middleware responds. One particular example was some researchers in Australia that really wanted to understand the feedback mechanism between the burning of the Savannah and the climate changes that cause monsoons. They used seven resources from the PRAGMA grid to run this simulation using software they developed for more than six months. They used more than a million hours of CPU. They were actually able to determine an outcome of science, under certain circumstances, that the burning of the Savannah does affect the onset of monsoons. This is a very interesting feedback and example of how we pooled resources and worked together in order to improve scientific results.
This is a list of other current applications that are using aspects of that distributed grid that we have and they range across different institutions using different technologies. In order to improve the management of these allocations, we work together and have created four groups (BeSTGrid from New Zealand, AIST from Japan, IHPC in Singapore and UCSD in the United States) to try to implement this software to make it easier, once we get these working groups started and these projects set up, for people to be added and become part of that working group. This exemplifies how different people come together to try to solve particular problems.
In the Biosciences, what we really want to do is focus on Avian Flu and see if there is a way of creating an infrastructure that will allow us to better set up virtual screening for this particular disease. And we draw upon resources from many different members of the PRAGMA community in this case. One particular algorithm that we draw upon from UCSD was developed by J. H. Chen who is now a post doc researcher here at Academia Sinica. And what it was the recognition that when people take a look at the docking of the molecule with the protein --that protein was rigid when it was crystallized—by default, that protein was moving around. And so we applied electro dynamics to the crystal to get different snapshots of that protein and then use that collection of snapshots to actually then do the docking across a library of compounds. And this approach has actually been used quite successfully in both the identification of HIV integrase. It’s also been used in sequencing as well as in researching the Avian Flu.
Other technologies that have been used have this involvement of (incomprehensible) University in China, that is, the use of their scheduling software, CSF4, that actually schedules jobs across different grid activities. We’ve coupled that with GFarm, developed at AIST, which is a Bio management system. And one of the things that has happened recently is they are trying to make the scheduling software more “data aware” because if you can move more closely to the data you can perhaps save time. As you can see in this particular diagram here, the more data that you have, the data aware plug-in is actually saving you quite a bit of time.