2.2A TECA an Extreme Event Detection and Climate Analysis Package for High Performance Computing

Thursday, 14 January 2016: 11:15 AM
Room 344 ( New Orleans Ernest N. Morial Convention Center)
Burlen Loring, LBL, Berkeley, CA; and S. Byna, M. Prabhat, J. Gu, H. Krishnan, M. Wehner, and O. Ruebel

TECA(Toolkit for Extreme Climate Analysis) is an open source extreme event detection and climate analysis package written in C++ designed to process massive climate data archives in parallel on modern supercomputing hardware. TECA is comprised of a set of applications built on its extensible framework which includes a set of core detection and analysis algorithms enabling climate scientists to harness the power of surpercomputing in their research. Support for scalable parallel design patterns is provided ranging from map reduce, to both task based and distributed data parallelism. The framework library includes a core set of climate detection and analysis algorithms including atmospheric river detection, tropical and extra tropical cyclone detection, CF I/O components, and supporting data manipulation algorithms. In this talk we will present the TECA architecture and discuss design decisions that have lead to demonstrated scalability to over 100000 cores on Mira, an IBM Blue Geen/Q system, and provide a highly extensible and easy to use product. We will present recent results obtained using the atmospheric river detector, and tropical and extra-tropical cyclone detectors in large scale analysis runs made at scale on Edison a Cray XC system.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner