The AppLeS TGV, using Cichlid as a visualization front-end.
The AppLeS TGV, using Cichlid as a visualization front-end.
GTOMO is an embarrassingly parallel tomography application being used in production at the National Center for Microscopy and Imaging Research (NCMIR). It uses a work queue scheduling strategy to reconstruct the 3-D structure of an object based on a series of 2-D images produced by NCMIR's electron microscope. GTOMO uses Globus services to run over a Grid environment. In order to achieve performance on the Grid, GTOMO is integrated with an application-level scheduler (AppLeS) which enables its tasks to be simultaneously scheduled over workstations and immediately available supercomputer nodes.
The above pictures display the GTOMO AppLeS when run on workstations available in the Parallel Computation Laboratory (PCL) at UCSD, workstations available at NCMIR, and SDSC's SP2. The green spheres represent processing nodes; the sphere turns white when processing is performed. A large green sphere is used to represent the SP2. The text attached to the spheres give the host name and the number of processors being used. The purple sphere represents the 'reader' node which sends tasks to available nodes for processing. The disk attached to the purple sphere represents the percentage of work left. Similarly the pink sphere represents the 'writer' node which receives processed data from the processing nodes. The disk attached to the pink sphere represents the percentage of work completed.
Image and text courtesy of Shava Smallen / UCSD-AppLeS. Used with permission. See the AppLeS Page for more info.