Project

CODAR: Center for Online Data Analysis and Reduction

Project Status: Active

Oak Ridge is an active participant in the Exascale Computing Project's (ECP) Codesign Center for Online Data Analysis and Reduction.  Ian Foster from Argonne is the lead PI, with Scott Klasky as the Oak Ridge site PI.

CODAR Logo

The CODAR project is lead by Ian Foster of Argonne National Laboratory.  The management team includes Todd Munson (ANL), Scott Klasky (ORNL), and Kerstin Kleese Van Dam (BNL).  

CODAR is focused on addressing the needs for data reduction, analysis, and management in the exascale era.  Given the significant increase in the amount of data that can be processed on such systems compared to the amount of data that can be written to the file system, investigators need to adopt new techniques for reducing the volume of data online, during the course of the simulation.  This can be in a coarse way, by simply increasing the number of time steps between writes, but for many types of simulation codes, this will reduce the quality of information that the end scientist receives.  To address this, CODAR evaluates, develops, and assists in the deployment of tools and techniques that allow for online reduction, compression, and/or analysis of the simulation output.

There are three corresponding thrusts or teams within CODAR -- Reduction, Analysis, and Infrastructure.  Oak Ridge participates strongly in all 3 teams.  In addition to these 3 core thrusts, there is an over-arching project focus on formulating and conducting codesign experiments to give scientists predictive characterizations of how different software, algorithm, and hardware choices impact performance, fidelity, and/or throughput.  The ADIOS project forms a key part of the software and codesign infrastructure.

For further information, please see ResearchGate project page, the Exascale Computing Project splash page, or the introductory HPCwire article.