Tera Scale Supernova Initiative A Networkers Challenge http

Need Boltzmann Solution ðNeed Angular Distribution ðNeed Spectrum ðNeed Neutrino Distribution Ê Fluid Instabilities

Dominant Computation: Nonlinear, integro-partial differential equations for the radiation distribution functions. Spherical Symmetry Axisymmetry

Data Management 3 D Hydrodynamics Run ð 5 Variables (Density, Entropy, Three Fluid Velocities)

Networking Bulk Data Transfer Needs Raw Bandwidth Needs for Collaborative Visualization Need end-to-end dedicated

Addressing Bulk Data Transfer Needs: Logistical Networking ð ð ð Light Weight Low Level

¯ Without putting in place the needed computational science infrastructure, our science will simply

Slides: 8

Download presentation

Tera. Scale Supernova Initiative: A Networker’s Challenge http: //www. phy. ornl. gov/tsi/ Explosions of Massive Stars Relevance: ðElement Production ðCosmic Laboratories ðDriving Application 11 Institution, 21 Investigator, 34 Person, Interdisciplinary Effort ðascertain the core collapse supernova mechanism(s) ðunderstand supernova phenomenology e. g. : (1) element synthesis, (2) neutrino, gravitational wave, and gamma ray signatures ð provide theoretical foundation in support of OS experimental facilities (RHIC, SNO, RIA, NUSEL) ðdevelop enabling technologies of relevance to many applications e. g. 3 D, multifrequency , precision radiation transport ðserve as testbed for development and integration of technologies in simulation “pipeline” e. g. data management, networking, data analysis, and visualization With ISIC and other collaborators: 77 people from 24 institutions involved.

Need Boltzmann Solution ðNeed Angular Distribution ðNeed Spectrum ðNeed Neutrino Distribution Ê Fluid Instabilities Ê Rotation Ê Magnetic Fields Need these to few percent accuracy! 6 D RMHD Problem!

Dominant Computation: Nonlinear, integro-partial differential equations for the radiation distribution functions. Spherical Symmetry Axisymmetry No Symmetry Example: Boltzmann transport equation for spherical symmetry.

Data Management 3 D Hydrodynamics Run ð 5 Variables (Density, Entropy, Three Fluid Velocities) ð 1024 X 1024 Cartesian Grid ð 1000 Time Steps 43 Terabyte Dataset “The flea on the tail on the dog…” Multidimensional Neutrino Data 13 Petabyte Dataset ~3 Petabyte Dataset. . . in weeks to months on a PF platform.

Networking Bulk Data Transfer Needs Raw Bandwidth Needs for Collaborative Visualization Need end-to-end dedicated paths/bandwidth, on demand. ðInteractive visualization. ðReal-time collaboration. Need protocols that provide this capability. None exist that will give 10 Gbps throughputs and stable control. What about the radiation field data? @ 3 PB! ðWork with Nagi Rao (ORNL).

Addressing Bulk Data Transfer Needs: Logistical Networking ð ð ð Light Weight Low Level Deployable ð New Paradigm Ü Integrate storage and networking. Ü Multi-source, multi-stream. … Solution Data transfer rates 200 -300 Mbps using TCP/IP! Limit set by ORNL firewall. Greater rates expected ðoutside firewall, ðother protocols (e. g. , Sabul). Direct impact on. TSI’s ability to do work! Atchley, Beck, and Moore (2003)

¯ Without putting in place the needed computational science infrastructure, our science will simply be inaccessible in the future. ¯ Significant progress has been made in the areas of ð linear solvers, ð performance analysis and optimization, ð data management and analysis, ð networking, ð and visualization. ¯ In particular, Logistical Networking has provided an easily deployable solution to our current bulk data transfer needs and has had a significant impact on TSI’s current ability to do science. ¯ TSI’s future data management and networking needs are daunting. TSI will generate hundreds of. Tera. Bytes of simulation data per simulation within the next two years. What then? ¯ Meeting these needs will require every new idea. Investment now in networking technologies will allow us to meet these needs.