Speaker
Description
Detailed analysis
We had to deal with some implementation issues in deploying the WISDOM docking application via our new intergrid backend on both EGEE and globus infrastructures. In terms of security, the voms proxy was used to access both infrastructures transparently. One specific problem that we have faced was that the autodock shell script developed in the WISDOM is to use the globus-url-copy to get all the files including executable, ligands and protein files necessary to run the autodock application on the grid. The script did not work on globus resources because no job delegation is permitted. To address this issue, we have developed the gridway plugin that takes charge of the staging-in of all the files necessary to launch the autodock script on the grid. Unlike the WISDOM original script, what our new script does is just to run the autodock executable code. The built-in LCG backend module and Ganga wrapper script, on the other hand, is supposed to take care of the staging of all the files.
Conclusions and Future Work
Thanks to the easy-to-use interface that Ganga provides, our approach to Grid interoperability using Ganga allows users to readily harness the full power of multiple grids to carry out a large-scale deployment of ligand-protein docking across EGEE and PRAGMA grid infrastructures. We expect to collaborate with the Ganga development team to integrate the plug-in modules that we have prototyped into the future release of Ganga.
Keywords
Ganga, Grid Interoperability, Gridway, Autodock
URL for further information
http://ganga.web.cern.ch/ganga/
Impact
We developed a plug-in module for the autodock application that allows users to deploy and run the WISDOM autodock code on both globus-based and glite-based grid infrastructures without having to know anything about grid interoperability issues relating to the two difference grid infrastructures. We also developed a globus backend plug-in module called a gridway backend designed to enable job submission to globus resources as well. We have also integrated an intergrid plug-in module into the Ganga. What the intergrid backend does is to determine where jobs will be submitted either to globus or glite nodes depending on the given load information. Last but not least, we have also developed a LigandSplitter module that supports the bulk submission of the same number of ligand-protein docking jobs as the number of ligand files described in the ligand directory.
