lecture 2gallen/presentations/southpadre_2006... · 2006-06-27 · lecture 2 grid applications dr...
TRANSCRIPT
Lecture 2Grid Applications
Dr Gabrielle Allen
Center for Computation & Technology
Department of Computer Science
Louisiana State University
Grid Summer Workshop June 26-30, 2006
Some CCT Grid Application Areas! Computational Chemistry
" GridChem: Building community deployment of chemistry and othercodes (www.gridchem.org)
! Petroleum Engineering
" UCoMS: Grid-enabling reservoir simulations, drilling technologies,integration with sensors and wireless networks, dynamic data drivenscenarios (www.ucoms.org)
! Black Hole Simulations
" Large scale simulations, novel scenarios for task farming, job migration,remote visualization and steering
! Coastal & Environmental Modeling
" SCOOP & DynaCode: Data workflows with coupled models, dynamicdata driven scenarios, metadata. (scoop.sura.org)
! High Speed Network Applications
" EnLIGHTened & LUCIFER: remote visualization, adaptive networks,co-scheduling, video conferencing.
Grand Challenge Problems
in Astrophysics
! Cosmology
! Black Hole andNeutron StarModels
! Supernovae
! AstronomicalDatabases
! GravitationalWave DataAnalysis
! Drive HEC &Grids
Gravitational Wave Physics
Observations Models
Analysis & Insight
Complex Simulations
! Requires incredible mix of technologies & expertise!
! Many scientific/engineering components
" Physics, astrophysics, CFD, engineering,...
! Many numerical algorithm components
" Finite difference? Finite volume? Finite elements?
" Elliptic equations: multigrid, Krylov subspace,...
" Mesh refinement
! Many different computational components
" Parallelism (HPF, MPI, PVM, ???)
" Multipatch
" Architecture (MPP, DSM, Vector, PC Clusters, FPGA, ???)
" I/O (generate TBs/simulation, checkpointing…)
" Visualization of all that comes out!
! New technologies
" Grid computing
" Steering, data archives
! Such work cuts across many disciplines, areas of CS…
Computational Science Needs
! Freely available, modular, portable and
manageable environment for collaboratively
developing parallel, high-performance multi-
dimensional simulations
! Developed for Numerical Relativity, but now
general framework for parallel computing
(CFD, astrophysics, climate modeling,
chemical eng, quantum gravity, …)
! Finite difference, AMR (Carpet, Samrai,
Grace), soon FE/FV, multipatch
! Active user and developer communities, main
development now at LSU and AEI.
! Open source, documentation, etc
Cactus Code
NSF Black Hole Grand
Challenge
! 8 US Institutions
! 5 years
! Attack colliding black holeproblem
Examples of Future of Science & Engineering
! Require Large Scale Simulations, beyondreach of any machine
! Require Large Geo-distributed Cross-Disciplinary Collaborations
! Require Grid Technologies, but not yetusing them!
NASA Neutron Star
Grand Challenge
! 5 US sites
! 3 years
! Colliding neutronstar problem
EU Astrophysics
Network
! 10 EU sites
! 3 years
! Continuing theseproblems
Grand Challenge Collaborations
! Community Driven" Distributed communities share resources
" Video Conferencing
" Virtual Collaborative Environments
! Data Driven" Remote access of huge data, data mining
" Eg. Gravitational wave analysis, particlephysics, astronomy
! Process/Simulation Driven" Complex Simulations of Science and
Engineering
" Task farming, resource brokering,distributed computations, workflow
! Remote visualization, steering,interaction
Typical scenario:Find remoteresources (task farm,distribute)
Launch jobs (static)
Visualize, collectresults
Prototypes and demos:need to move to:
Fault tolerance
Robustness
Scaling
Easy to use
Complete solutions
Current Grid Application Types
! Addressing large, complex, multidisciplinary
problems with collaborative teams of varied
researchers ...
! Code/User/Infrastructure should be aware of
environment
" Discover and monitor resources available NOW
" What is my allocation on these resources?
" What is bandwidth/latency Code/User/Infrastructure
should make decisions
" Slow part of simulation can run independently …
spawn it off!
" New powerful resources just became available …
migrate there!
" Machine went down … reconfigure and recover!
" Need more memory (or less!), get by adding
(dropping) machines!
Dynamically provisionand use new highend resources and
networks
New Paradigms for Dynamic Grids
S1 S2
P1
P2
S1S2
P2P1
S
We see something,
but too weak.Please simulate
to enhance signal!
Future Dynamic Grid Computing
Found a black hole,Load new component
Look forhorizon
Calculate/OutputGrav. Waves
Calculate/OutputInvariants
Find bestresources
Free CPUs!!
NCSA
SDSC
RZG
LRZArchive data
SDSC
Add more resources
Clone job withsteered parameter
Queue time over, find new machine
FurtherCalculations
AEI
Archive to LIGOexperiment
Future Dynamic Grid Computing
New Grid Scenarios! Intelligent Parameter Surveys, speculative computing, monte carlo
! Dynamic Staging: move to faster/cheaper/bigger machine
! Multiple Universe: create clone to investigate steered parameter
! Automatic Component Loading: needs of process change,
discover/load/execute new calc. component on approp. machine
! Automatic Convergence Testing
! Look Ahead: spawn off and run coarser resolution to predict likely future
! Spawn Independent/Asynchronous Tasks: send to cheaper machine, main
simulation carries on
! Routine Profiling: best machine/queue, choose resolution parameters
based on queue
! Dynamic Load Balancing: inhomogeneous loads, multiple grids
! Inject dynamically acquired data
! Abstract programming
interface between
applications and Grid
services
! Designed for applications
(move file, run remote
task, migrate, write to
remote file)
! Led to GGF Simple API
for Grid Applications
(SAGA)
www.gridlab.org/GAT
Main resultfrom GridLab
project
Scp, DRMAA, Condor, SGE, SRB, Curl, RFT.Under Develop
Basic functionality, will work on single isolatedmachine (e.g. cp, fork/exec)
Default Adaptors
GRMS, Mercury, Delphoi, iGridGridLab Adaptors
Core Globus functionality: GRAM, MDS, GT-RLS,GridFTP
Globus Adaptors
Grid Application Toolkit (GAT)
Distributed Computation! Issues
" Bandwidth (increasing fasterthan CPU)
" Latency
" Communication needs,Topology
" Communication/computation
! Techniques to be developed" Overlapping
communication/computation
" Extra ghost zones to reducelatency
" Compression
" Algorithms to do this forscientist
Harnessing MultipleComputers
Why do this?Capacity: computers can’t keep upwith needsThroughput: combine resources
HTTP
Streaming HDF5Autodownsample
Any Viz Client:
LCA Vision, OpenDX
Changing steerableparameters• Parameters• Physics, algorithms• Performance
Remote Viz & Steering
Cactus Worm (SC2000)
! Cactus simulation starts, launched from
portal
! Migrates itself to another site
" Grid technologies
! Registers new location
! User tracks/steers, using HTTP, streaming
data, etc…
! Continues around Europe…
User only has to invokeCactus “Spawner” thorn…
Appropriate analysis tasksspawned automatically tofree resources worldwide
Task Spawning (SC2001)Cactus “Spawner” thorn automatically
prepares analysis tasks for spawning
Grid technologies find resources,
manage tasks, collect data
Intelligence to decide when to spawn
SC2001: resources of GGTC testbed.
Main Cactus BHsimulation starts
here
! 5 continents and over14 countries.
! Around 70 machines,7500+ processors
! Many hardware types,including PS2, IA32,IA64, MIPS,
! Many OSs, includingLinux, Irix, AIX, OSF,True64, Solaris, Hitachi
! Many organizations:DOE, NSF, MPG,universities, vendors
! All ran same Gridinfrastructure, and usedfor differentapplications
Cactus black holesimulations spawnedapparent horizon
finding tasks acrossthe grid.
Supercomputing 2001
Prizes for mostheterogeneous and most
distributed testbed
Global Grid Testbed Collaboration
Main Cactus BHSimulation startedin California
Dozens of low resolutionjobs test corotationparameter
Huge jobgeneratesremote datavisualized inBaltimore
Error measurereturned
Black holeserver controlstasks and steersmain job
Black Hole Task Farming (SC2002)
Job MigrationGridLab demonstration
SC2003
GridSpherePortal
SMSServer
Mail Server
“The Grid”
ReplicaCatalog
User details,notification prefs andsimulation information
IM Server
Notification & Information
Computational Chemistry Grid (CCG)
! NSF funded (LSU, UK,
OSC, TACC, NCSA)
! Aim to provide access to
computational chemistry
software, hardware, and
services using well
established Grid
technologies
! For expert, novice and non-
traditional users
! Gaussian, GAMESS,
NWChem, MolPro, + …
Terminology
! Computational Chemistry Grid: A virtual
organization that will provide access to distributed
high performance resources for the computational
chemistry community with distributed support
and services, intuitive interfaces, and a
measurable quality of service.
! GridChem: Java based desktop client
! Grid Middleware Server: Interfaces between
client and HPC resource, Handles authentication
and certificate management
! Coordinated use of compute and network resources
! Define the network through the set of services that it can offer
• MCNC (lead),
LSU, RENCI,
NCSU, Cisco,
AT&T, Calient
Networks
• 10Gig capacity
testbed running
over LONI and
NLR
iGrid 2005
Distributed & collaborative
visualization using optical
networks
! High-definition video
conference connecting
multiple sites
! Central visualization
server
! Interaction from all
participating sites
! Data on remote machines
! “Ubiquitous Computing &Monitoring System for Discovery &Management of Energy Resources”
! DOE/Louisiana BOR funded
" Petroleum engineering
" Wireless sensor networks
" Grid technologies
! Applications
" Reservoir simulation! Uncertainty analysis, sensitivity
studies, history matching
" Real-time well surveillance
" Drilling performance analysis withhigh-rate data
“UCOMS”
Oil Industry in Louisiana
Major oil producingstate in US:" 5th in production
" 8th in reserves
" Home to 2 of 4 strategicpetroleum reserves
" 17 petroleum refineries(capacity 2.8Mbarrels/day)
" Ports receive ultra largeoil tankers
" 20,000 oil producingwells, around 4Koffshore.
Reservoir Studies
! Assessments and predictions of oil/gas reservoir
performance, depending on
" Geological heterogeneity
" Engineering choices
! Used for development and operational decisions
… models assess different production scenarios.
! Applications:
" Well placement & performance prediction
" Sensitivity analysis & uncertainty assessment
" History matching (model verification and tuning)
Reservoir Simulation
! Mathematical model for fluid flow in a reservoir involvesdensity, permeability (K), mobility, pressure (P), productionrate (q), porosity & saturation, where m denotes either oil,water or gas.
! Many geological parameters cannot bemeasured or modeled and are unknowns.
! We are using UTChem (3D, multiphase,multicomponent, compositional, variabletemperature, FD simulator)
Core Computational NeedsCompute:
! Large-scale computation: Seismic inversion, flow numericalsimulations, Ggeostatistics, analysis, …
Data:
! Large data sets (TB-PB): Seismic, Geological & Geophysical(G&G), Well logging, Simulation results, …
Security:
! Commercial benefits lead to high security for all data, proprietarycodes, etc.
Workflow:
! Parameter selection, model construction, data movement, modeldeployment, results analysis etc.
Move towards dynamic, data driven scenarios, including direct inputfrom sensor data.
Current Workflow
! Example case: Eleven geological factors e.g. initial pressure+ three engineering factors e.g. tubing diameter with either 3or 4 levels.
! Factorial design:" 4^6 x 3^8 = 26,873,856 reservoir simulations
" 100 days on 1024 proc cluster (at 6 mins per run)
! Even with experimental design many runs needed
Motivation for ResGrid Infrastructure
! Reservoir simulation is one of the largest users ofcomputing power
" Large, complex, uncertain models
" High risks and rewards
! Can be performed more efficiently if
" Moderate-sized jobs can be farmed out onto aheterogeneous, underutilized grid
" Large jobs can be run in parallel on a grid
! Efficiency gains can be used to assess risks,estimate parameters, run larger and more complexmodels, and optimize developments
ResGrid Usage Scenario
! Five key components:
" Portal
" Reservoir modeling
" Large scale distributedreservoir simulation
" Analysis
" Visualization
! Light weight client to
locate and replicate data
files
Data Replication Tool: getdata
! SuraGrid is a beyond-regionalinitiative in support of regionalstrategy." Grid Infrastructure connecting states in
South-Eastern US.
" conceived to enable seamless sharingof resources between variousinstitutions to further scientificresearch and Grid development.
! Looking for applications !!!
! Partnering with IBM and othervendors to procure HPC Resourcesacross various SURA sites to build astrong computational resourcebackbone.
! Many sites connected via high speedoptical networks like NLR and LONI.
! http://www.sura.org/suragrid
SURAgrid
Louisiana Coastal Area (LCA)! 1927 flood, levees, loss of wetlands,
growing crisis, social impact
" 25% lost wetlands in last century, futurepredictions dire; increases flooding, surgerisk
! Hurricanes Katrina, Rita, Wilma (2005)
" 1.4M FEMA aid applications, 35K > 1000miles away, 33K evacuations by coastguard alone (6x 2004). Years to rebuild
! Important problems: ecological,hurricane, algal bloom/salinityforecasting, restoration, evacuation,emergency response strategies
! Rich dynamic environment for modeling:coupled models, multi-scale, realtimedata (sensors, satellites)
! Important role for HPC, Models, Grids,Community
Holly Beach (Hurricane Rita)
Holly Beach (Hurricane Rita)
North Breton Island, La.
looking Northwest.
South and East of Mouth
of Mississippi River
After Katrina
Beyond Katrina: Land Loss
Katrina Modeling: Tracks
Katrina Modeling: Storm Surges
! Storm surges: the worst part ~ 25ft for Katrina, kill far more peoplethan winds
! ADCIRC: Joannes Westerink,Rick Luettich, Randy Kolar, ClintDawson
! Input 2D Unstructured Mesh, wind,pressure
" 314K nodes, 85% near LA Coast
" 50km resolution in deep ocean,100m resolution
! 128 processor supercomputer: 1hour to forecast 1 day
! Want: run dozens of simulations
" Vary inputs, paths, strength
Source: http://www.nd.edu/~adcirc/
Computational domain withbathymetry (m)
Unstructured grid of the entiredomain.
ADCIRC Computational Mesh
Model-Model-Data Coupling
SURA Coastal Ocean Observing Program (SCOOP)
! Integrating data from regional observingsystems for realtime coastal forecasts in SE
! Coastal modelers working closely withcomputer scientists to couple models, providedata solutions, deploy ensembles of modelson the Grid, assemble realtime results withGIS technologies.
WAVCIS:Wave-Current-Surge Information System for Coastal Louisiana
! Program Director: Greg Stone, CSI, LSU
! Provide wave information (sea state)including wave height, period, direction ofpropagation, water level, surge, nearsurface current speed and direction andmeteorological conditions on a real timebasis around the entire Louisiana coast.
DynaCode: DDDAS framework for coast and
environmental modeling
! Builds on SCOOP modeling expertise
! LSU (CCT, CS, LHC, CSI, WBI), Notre
Dame (+ EU DROIDS)
! Two complimentary scenarios:
– Coupling ocean circulation, storm surge,
wave generation models for the Gulf
– Coupling ecological, hydrodynamic,
sediment transport models of the Mississippi
River Delta
! Infrastructure & algorithms
– Couple multi-models, data, external inputs
from sensors, wind & databases
– Optimize complex workflows on grids,
invoking appropriate models, meshes,
algorithms, depending on conditions.
DynaCode: DDDAS framework for coast and
environmental modeling
! New capabilities:
" dynamically invoke more accuratemodels and algorithms as hurricaneapproaches coast,
" choose appropriate computingresources for needed confidencelevels
" compare model results withobservations to feedback into runningsimulations
" realtime data assimilation
" adaptive multi-scale simulations
" dynamic component recomposition
" simulation needs steer sensors anddata input
Cyberinfrastructure
High End Computing
Computer Science
Applied Math
Application Domains
DynaCode! Focus on scenarios:
– Hurricane ensemble modeling
• Coupling ocean circulation, storm surge,wave generation models for the Gulf
• Notifications from NHC trigger customizedensemble hurricane models(surge/wind/wave), sensors verify, guidedynamic ensembles
• Event driven, dynamic componentframework with algorithm selection,optimization tools, workflow, dataassimilation, result validation withsensor/satellite.
– Ecological restoration and control
• Breton Sound diversion, control structureto allow Mississippi to flow into wetlands
• Coupled models (hydrodynamic, salinity,geomorphic, sediment) control diversion,sensors/wind fields inject real time data.
! SC05 Streaming
HD Video
" Starlight/Chicago to
SC showfloor
" CaveWave/NLR
optical network
" Data size - 21G
" Bandwidth -
500Mbps
" Frame-rate - 12fpsHigh speed networks critical;
Collaborative technology brings
experts together across state & world
Want Real Time! LONI! NLR!
Emergency Response
! Run many simulations if possible
" High res, high throughput critical: HPC + Grids!
" Presently 6 hours to convert NHC forecast to surge
! 240 processors per run
" Bracket official forecasts with different tracks
" Create “MEOW” map
! Notify officials: FEMA, State, Governor’s Office
" email, web, phone: can do better!
" evacuation orders: based on this work!
! Continue forecast to inform emergency workers
Emergency Forecast
Emergency Forecast
Very accurate forecasts
are possible;
reliability can be improved
Used to notify officials,
FEMA, State,
Governor. Email, web,
phone: can do better!
Components
! Infrastructure & algorithms to couple models, to eachother and to external inputs from sensors, wind &databases to optimize execution of complex workflows ongrids, invoking appropriate models, meshes, andalgorithms, depending on current conditions." Applications: control algorithms and coupling interfaces for
coastal/eco-codes
" Maths: Model errors for scenarios and implement basic statstoolkit in Cactus to drive ensembles
" Systems: Enhance dynamical capabilities of Cactus/Triana,decision making infrastructure, dynamic recomposition. Addlegacy ensemble modeling to Cactus. Think about tracking dataflow/data sensitivity through Cactus.
" Measurements: Integrate data to scenarios from existing sensors,plan out interfaces for sensor control.
! $500M to restore 250 sq. miles since 1986
– 12 sq. miles/year, but losing 2x this per year!
! How to catch up? $14B dollars later...
– floods? diversions? pipelines of mud? many ideas!
! Complex processes: comprehensive approach needed to understand
competing forces. But there is a quantitative answer!
Wetland Restoration
Changing How Science is Done
Source: Building a Cyberinfrastructure for the Biological Sciences, NSF Report 2003
Cyberinfrastructure
Applications• Environmental sciences• Computational fluid dynamics• Bioinformatics• …
Hardware• Compute resources• Networks• Storage• …
Programming & Developmenttools & libraries
Services & Middleware
Domain specific software tools
Sharedsoftware tools
Distributed hardware
}
Disco
very
& I
nnov
ation
Educ
ation
& T
raining
Integrated Cyberinfrastructure
Finally …
! Any questions … ask Archit over the next days
! Links from my web pages
http://www.cct.lsu.edu/~gallen or CCT Web
pages http://www.cct.lsu.edu
! Email me … [email protected]
! Very interested in talking to people interested in
using Cactus, GridSphere, GAT/SAGA, working
with SURAgrid
! Visitor program at CCT, setting up student
internships.