Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Robert Jacob Project Name: parvis Division: MCS Project title: Parallel Analysis Tools for Ultra-Large Climate Data Sets Associated funding: DOE BER Other Systems: NCAR, NERSC, ALCF, OLCF Science: Like many scientific applications, climate models rely on numerical methods to solve their equations. But insights about the simulated climate are revealed only after performing additional calculations on several years to decades of the multivariate, multidimensional time series output by the climate model. The tools currently used to perform the climate data analysis and visualization (DAV) are serial and can no longer handle the data sizes being generated. We will be developing a new parallel version of a widely used climate DAV system called NCL which will speed up the climate science workflow for researchers throughout the community. Project description: NCL (the NCAR Command Language) is an interpreted scripting language that allows Matlab-like operations on arrays, can read many climate and weather data formats and provides special functions for plotting climate and weather quantities over maps of the earth. NCL produces publication-quality output. To parallelize NCL, we will use two existing parallel libraries. MOAB, the Mesh Oriented Database, contains datatypes and methods for representing unstructured grids in parallel. Intrepid is part of the Trilinos system and is a library for computing operators acting on discretized fields. MOAB and Intrepid both use MPI for their parallelism and MOAB, Intrepid and NCL are written in C++. We will be developing new code to interface NCL with MOAB/Intrepid. Along with developing a parallel version of NCL, we are also investigating the use of task parallelism using the Swift language. We are converting current c-shell scripts that analyze climate data into Swift scripts that can perform the steps in parallel. We are also substituting some of the serial commands in the scripts with parallel versions so the resulting script will be both task- and data- parallel. Are goal is for a scalable DAV tool that can perform all I/O and calculation in parallel while still rending publication quality 2D plots. If we distribute the data evenly among processors in parallel, then all time-domain calculation will be embarrassingly parallel while most spatial calculation will need only nearest neighbor exchanges. Our requested time will be use to test and time the scaling of our software as it is developed. Project URL: http://trac.mcs.anl.gov/projects/parvis Current FY Hours Used: undetermined amount New FY Requested allocation: 4000 Q1: 1000 Q2: 1000 Q3: 1000 Q4: 1000 Justification: Thank You, The LCRC Accounts System