neesgrid: where infrastructure meets cyberinfrastructure

18
NEESgrid: Where Infrastructure Meets Cyberinfrastructure Kim Mish Presidential Professor of Structural Engineering Director, Donald G. Fears Structural Engineering Laboratory School of Civil Engineering and Environmental Science University of Oklahoma

Upload: calum

Post on 15-Jan-2016

23 views

Category:

Documents


0 download

DESCRIPTION

NEESgrid: Where Infrastructure Meets Cyberinfrastructure. Kim Mish Presidential Professor of Structural Engineering Director, Donald G. Fears Structural Engineering Laboratory School of Civil Engineering and Environmental Science University of Oklahoma. So Why am I Here, Anyway?. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

NEESgrid: Where Infrastructure Meets Cyberinfrastructure

Kim MishPresidential Professor of Structural Engineering

Director, Donald G. Fears Structural Engineering Laboratory

School of Civil Engineering and Environmental Science

University of Oklahoma

Page 2: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

So Why am I Here, Anyway?

• Professional Background– Earthquake Engineering (structures, bridges, dams, infrastructure)

– Information Technology and Supercomputing

– National Security Research and Development

• Established LLNL Center for Computational Engineering– Interfaces of Simulation, IT, and INFOSEC for LLNL Engineering

– Substantial university and government outreach component

• Currently providing technology management expertise for the NSF NEES MRE– Primary focus has been on the NEESgrid project (SI award)

– This project lies at interface of infrastructure and cyberinfrastructure realms, a.k.a. where I’ve spent my career

Page 3: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Characteristics of Infrastructure

• Essential– So important that it becomes ubiquitous

• Reliable– Example: the built environment of the Roman Empire

• Expensive– Nothing succeeds like excess (e.g., Interstate system)

– Inherently one-off (often, few economies of scale)

• Clear factorization between research and practice– Generally, only deploy what provably works

Page 4: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Infrastructure vs. Cyberinfrastructure

• Characteristics of Infrastructure Culture– Risk averse, which leads to slow technology adoption– Code-based practice to defend against litigation– Follow community wants/needs whenever possible– Goal is highest reliability, e.g., MTBF

• Characteristics of Cyberinfrastructure Culture– High-risk, “innovate or die” approach to technology– Best-practices approach leaves legal issues dangling– Develop technology, then look for a market– Goal is highest performance, e.g., TFLOPS

• Two communities with nothing in common!

Page 5: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

So Exactly What is NEES?

• NEES = Network for Earthquake Engineering Simulation– NEES is a distributed array of experimental sites, grid-based data

repositories, tool archives, and computational resources

• NEES has four components:– The consortium, which has run NEES since late 2004– The consortium development effort, which built the consortium– The experimental sites, which provide data and content– The systems integration (SI) effort, termed NEESgrid

• IT drivers include telepresence, curated repositories, scalable HPC, experimental-numerical coupling, QoS…

• NEES is the first-ever Engineering MRE at NSF, and its full title is the “George E. Brown, Jr. Network for Earthquake Engineering Simulation”

Page 6: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

NEES: Experiments and Numerics

• Network for Earthquake Engineering Simulation

Goal: create collaborative network of experimental sites

at fifteen U.S. universities

Page 7: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

NEES: UC Davis Soil Centrifuge

Page 8: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Oregon State Tsunami Facility

Page 9: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

What About Numeric Simulation?

• GC Example: San Francisco-Oakland Bay Bridge– Horrendous nonlinearities and ill-conditioning– Foundation is saturated (fully-coupled multiphysics)– Complexity from juxtaposition of forms (unstructured)

Page 10: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

A Closer Look at the Bay Bridge

Superstructure:Ill-conditioned numericsMaterial nonlinearitiesGeometric nonlinearities

Substructure:All the problems of the superstructure AND of

the foundation

Foundation:Material nonlinearities

Coupled fluid-solid multiphysicsImpossible numerics

Free-Field Response:Simulation has spatial limits, but the physical problem doesn’t.

Validation and Verification: Don’t even ask!

Page 11: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Analysis of Site, Dam, and Reservoir

• Interoperability: NIKE3D/DYNA3D data

Page 12: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Morrow Point Finite-Element Mesh

• Analyze foundation, dam, and fluid in lake

Page 13: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

The Promise of Cyberinfrastructure• NEESgrid Example: the

Terascale Framework– New client-server

engineering portal for grid computing

– Scalable framework for finite-element HPC

– Developed by Lee Taylor (SNL ASCI flagship SIERRA framework lead)

– Funded by LLNL CCE, SNL, and NSF ITR in support of NEES MRE

Page 14: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

The Perils of Cyberinfrastructure

Page 15: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

NEESgrid Architecture Problem

• Diverse user communities & applications– 10s of experiment sites, 100s of user sites,

1000s of users (or more, eventually)– Access to data, simulation, collaboration, etc.

• Demanding performance requirements– Response time, data volumes, security, scale

• Impractical to meet these requirements with non-engineered stove-pipe solutions

Page 16: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

What Went Wrong with NEESgrid

• Designed from bottom-up with virtually no requirements gained from users– Technology-push almost never works!– Grid developers (Globus) had no idea how to

deliver production software component

• Useful software promised always, delivered almost never, users got fed up with the wait– Does this sound familiar (Multics)?

Page 17: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

How We Develop Infrastructure

• Multi-tiered structure for R&D– NSF: basic engineering research– TRB: development and reduction-to-practice– FHWA, AASHTO, and DOTs: deployment of

innovations that are successful and feasible

• Clear lines of demarcation exist– Don’t do research on production facilities– Use funds from production to support R&D

• When in doubt, overbuild!

Page 18: NEESgrid: Where Infrastructure Meets Cyberinfrastructure

FearsLab Donald G. Fears Structures LaboratoryUniversity of Oklahoma

Summary

• Design and deployment of infrastructure is motivated by the goal of production capability with low risk and high reliability

• Design and deployment of cyber-infrastructure is motivated by the goal of performance and technological innovation

• The NEES MRE lies at the oft-problematic interface of these two communities