About

Created & Maintained for You By

Auton Universal Viewer was created for you by Gus Welter, Anthony Wertz, and Dr. Artur Dubrawski of the Auton Lab at Carnegie Mellon University.

Current contributors & maintainers include Roman Kaufman, Vedant Sanil, Stefania La Vattiata, and Gus Welter.

Additional Contributors

Additional contributors & collaborators include:

Dr. Michael R. Pinsky, Critical Care Medicine, University of Pittsburgh Medical Center
Dr. Gilles Clermont, Critical Care Medicine, University of Pittsburgh Medical Center
Dr. Marilyn Hravnak, University of Pittsburgh School of Nursing
Victor Leonard, Auton Lab, Carnegie Mellon University
Many others who provided valuable input & feedback.

Interested in Contributing?

Are you interested in contributing to Auton Universal Viewer? Please see Contributing to Auton Universal Viewer

Design Choices

AUViewer is designed to enable web-based viewing & annotation of large time-series data via a web browser and across the internet in a portable way. To unpack some of these terms:

large time-series data: E.g. 20-100GB per file and beyond. File size can theoretically grow without limit and still remain performant for a user in a browser on a standard internet connection.
across the internet: Using a basic internet connection, e.g. 1 Mbps download
portable: Many software packages for working with large datasets require software installation. AUViewer was made to handle such tasks with no software installation, special compute resources, or special bandwidth required of the end user.

Tradeoffs & Implications

Any non-trivial system involves design choices and tradeoff decisions. AUViewer prioritizes performance and memory efficiency over disk efficiency.

Specifically, in order to render large time-series datasets in the browser quickly and without significant backend memory requirements, the system pre-computes and stores to disk min-max downsample representations of all original files which it serves. Generally, this leads to a doubling of original data disk space requirements (e.g. a 20GB original file will require on the order of 20GB of additional space for pre-computed downsample data).