GridMeetingMay2108

Summary

This was the second meeting of the FLR grid group. Technical representatives of University of Buffalo's Center for Computational Research (CCR) were also invited to participate. Their presence was welcomed both on account of their experience dealing with similar issues in operating NYSGrid and because both projects recognize the advantages of sharing effort and resources. Their project uses several Open Science Grid mechanisms.

Two issues were discussed: (1) determining the details of running the FGF application (FgfVascularization) that was selected as the prototype for execution on Florida HPC resources, and (2) exploration of possible ways to implement authentication for users running on Florida resources.

FGF application: Jim W. already provided a nice description of the FGF application (FgfVascularization). He believes that FDF is relatively straightforward and should run on foreign resources without a problem. Of course, we want to verify this by actually running it on several HPC sites in Florida without any grid authentication mechanisms (this means that an account must be set up on every resource where it will execute). Sites wanting to provide resources should go to the FGF page and add their information at the bottom of the page, following the FSU example which Jim inserted today.The question was raised about whether FGF would run on IBM p5 resources in Miami.

Authentication, documentation, support, effort: (This section is unfinished) This discussion is difficult to summarize concisely because it ranged over a variety of issues, including authentication mechanisms, complications from multiple VOs (virtual organizations), user support, the FloridaGridOperationsCenter (fGOC), OSG mechanisms, documentation and effort. There was general agreement that we should set up a "Florida VO" with certificates issued by a local CA so that Florida users could be authenticated without requiring the more stringent DOEGrids certificates that are accepted at all OSG sites. Authentication would probably be required from the user's home institution. How would we do this?

All agreed that finding the necessary effort to operate FLGrid will be challenging but necessary. A guess was made that approximately 0.25 FTE would allow participation in minimal FLGrid activities, but user support could easily soak up additional manpower. However, getting extra human resources will require an active effort. Jim W. and Paul A. agreed to look at the possibility of writing a brief document summarizing FLGrid's vision and existing resource sharing (we need to get a few users to make the case effectively) and requesting modest support at our institutions. Even Buffalo seems to have minimal extra resources devoted to their effort (Question: did CCR get an additional 0.25-0.50 FTE to help?)

We might be able to use Paul A.'s student who helps provide documentation for the OSG Grid training effort. He might be able to adapt some existing documentation on procedures and post it on the FLGrid TWiki. However, we need a plan before we can document it. There is also the possibility of utilizing some effort from a new Tier-2 member whose time is split between OSG/CMS and a Florida-based education/physics/network project called CHEPREO, which already has a grid component.

Participants

• Paul Avery (UF)
• Jim Wilgenbusch (FSU)
• Paul Van der Mark (FSU)
• Dan Majchrzak (USF)
• Jorge Rodriguez (FIU)
• Jon Akers (UF)
• Dave Pokorney (UF)
• Steve Gallo (U Buffalo)
• Matt Jones (U Buffalo)