Website header

Twister visualization server deployment

Toroidal magnetic fields generated by dynamo action in the convection zone of a star like our sun, but rotating more rapidly at three times the current solar rate, as our own sun did when younger. This image was produced on twister, CISL's prototype TeraGrid visualization resource, by Ben Brown, University of Colorado. In FY2007 a visualization computer named "twister" was procured, security hardened, and deployed as a prototype TeraGrid Data Analysis and Visualization (DAV) node. The system was a prototype in the sense that it was not a fully integrated TeraGrid resource (e.g. lacking the CTSS software stack), used primarily for GPFS WAN and remote visualization evaluation purposes, and available only to a limited number of users. Nevertheless, throughout much of FY2007 and FY2008 twister provided VAPOR-based visualization services to researchers at the University of Colorado who were exploring terabytes of astrophysical fluid flow simulation data computed and stored at SDSC. These data, outputs from the Anelastic Spherical Harmonic (ASH) model, were mounted at NCAR using GPFS-WAN and analyzed remotely from CU via HP's Remote Graphics Service.

In FY2008 CISL began the process of fully integrating twister into the TeraGrid as a production DAV resource. The integration process has been somewhat onerous and time consuming, as the TeraGrid lacks provisions (e.g. templates, scripts, or documentation) to facilitate standing up new resources. Integration of twister is nearing completion, and twister is expected to be announced as a new TeraGrid resource at end-FY2008.

Twister, purchased primarily to enable proof-of-concept work, offers only meager computation resources, capable of supporting only a very small number of simultaneous users. In FY2009, assuming utilization warrants, we will acquire and deploy a more capable DAV system. We will also deploy a resource scheduling mechanism suitable for interactive use, an activity that dovetails with broader CISL plans.

This work is supported by NSF Core funding.