Revolution Enabling Large-Scale Collaborative Science.
Post on 25-Dec-2015
216 Views
Preview:
Transcript
Revolution
Enabling Large-Scale Collaborative Science
Outline
Introduction Visualization Applications Distributed, Parallel, Grid-based, and Collaborative
Visualization Collaborative Scientific Visualization Environments (CSVE) Future Directions
Revolution In Science
Pre-Internet• Theorize &/or experiment, alone or in small teams; publish paper.
Post-Internet• Construct and mine large databases of observational or simulation data.• Develop simulations, analyses, & synthesis.• Access specialized devices remotely.• Exchange information within multidisciplinary teams.
Image from CERN
Image from www.aip.org
Why Visualization?
Visualization is now seen as an integral part of modern computing
High performance computing generates vast quantities of data
High resolution measurement technologies generate vast quantities of data
Information systems incorporate large data sets and complex relations
We simply must harness our visual systems to aid us in understanding our data
What Is Visualization?
Medical Applications
From MRI, CT, Confocal Microscopes, …
We can visualize human anatomy at various scales
Curved Surface through the aorta tree. Visible Human Server, from R.D. Hersch at Ecole Polytechnique Fédérale de Lausanne
Optical nerve in the retina. Imaris software from B. Ehinger, Department of Ophthalmology, Lund University Hospital Torn ACL. Anonymous image.
Climate Applications
From simulators, satellites, measurement stations, …
We can visualize events, climate, and current weather
Satellite and surface image for January 19, 2004. Image from Unisys Weather.
These images show a comparison between two large El Niño events. The first begins in Oct '81 and the second in Oct '96. Image from NCAR.
Top-south view of 3-D volume of the simulated Andrew's radar reflectivity. Image from Y. Liu, McGill University.
Oil and Gas Applications
From simulators, seismic data sets, field measurements, …
We can visualize production, management, and exploration
Streamlines emanating from a virtual well show a three-dimensional oil flux. Image from Lawrence Berkeley National Laboratory.
Immersive visualization of horizons, faults, wells, and salt dome. Image from BP Visualization Center, University of Colorado.
Real-time cross-section planes where opacity is reduced in order to show values of interest. Image from HueSpace of Norway.
Molecular Applications
From simulators, experiments, measurements, …
We can visualize molecules, simulated values, and statistical measurements
The image depicts the electrostatic potential at each point of the Van der Waal's dot surface around aspirin. Image from Roger Sayle
Main chain hydrogen bonds and peptide bonds deviating more than some degree from planarityImage from Dirk Walther, UCSF.
Fancy CPK model. Atoms are made of various metals (C: gold, H: chrome, N: bronze, O: silver, S: brass). The ellipsoid (made of red glass) is the one with the smallest volume containing 70% of all atoms. The molecule is Trypsin Inhibitor. Image from L. Chiche .
Environmental Applications
From observations, experiments, measurements, …
We can visualize terrain, database information, and measurements
Populations of trees using a range of rendering techniques. Image from USDA Forest Service, Pacific Northwest Research Station .
Monitoring wind profile in Monterey Bay. Image from A. Pang, UCSC.
Patterns of recent forest management activities in the Northwest. Image from J. S. Nighbert, Oregon BLM.
Scientific Visualization
1987 NSF Report B.H. McCormick, T.A. DeFanti, and M.D. Brown, "Visualization in Scientific Computing," in Computer Graphics, Vol. 21, No. 6, (special issue).• Turning “firehoses” of data into a
visual representation• Enabling the scientist to “see the
unseen” Argued that investment in high
performance computing in US was wasted unless there was corresponding investment in visualization
Led to the development of several visualization software systems
One of the many visualization software systems created during this time. Developed by B. Hibbard. vis5d.sourceforge.net
Dataflow Visualization
Visualization represented as a pipeline:• Read data• Filter data• Map data• Render data• Display data
System realized in at least two ways:• Modular Visualization Environment• Toolkits or Libraries
Modular Visualization Environment
Modular Visualization Environments• IRIS Explorer, OpenDX, AVS, …• Visual programming paradigm -
allows easy experimentation which is what one needs in visualization
• Extensible – add your own modules
• Scientist uses ‘visual programming’ to connect modules together
AVS5 – www.avs.com
OpenDX – www.opendx.org
IRIS Explorer – www.nag.com
Visualization Libraries and Toolkits
Visualization libraries and toolkits OpenGL, Java3D, VTK, OpenRM,
Java3D, …• Provides the application programmer
an API• Scientist uses applications or
incorporates visualization code in own software
• Open source OpenGL
• Industry standard• Hardware acceleration• Basis for VTK, OpenRM, Java3D
Java3D• A mapping of OpenGL
OpenRM• Direct volume rendering
VTK• Bindings to Tcl, Python, Java
Virtual creatures from Stanford University using VTK – www.vtk.org
Visapult from LBNL usingOpenRM – www.openrm.org
vmd from NCSA using OpenGL – www.opengl.org
Cave using Java3D from the University of Calgary Java3D – java.sun.com/products/java-media/3D/
Visualization and Simulation
Visualization is a key tool in understanding the results of numerical simulations of complex phenomena
Use cases of visualization for simulation
• Pre-processing• Treat dataflow visualization
environment and simulation as separate activities
• Tracking• Replace data in visualization
pipeline with the simulation• Track behavior
• Steering• Include control module in
visualization pipeline• Simulation responds to
visualization environment• Post-processing
• Again, treat visualization and simulation as separate activities
Pre-process
Track
Steer
Post-process
Reservoir simulation using VTK from Geocap
Visualization and Observation
Visualization is a key tool in understanding observational data
Use cases of visualization for observational data• Monitor
• Monitor incoming observations
• Post-processing
• Treat visualization and observations as separate activities
• Integration
• Accept multiple input streams
Monitor
Integrate
Post-process
Monteray Bay monitoring from REINAS, UCSC
Distributed Visualization
Distributed visualization• Offload some computationally
intensive tasks• Couple the simulation with the
visualization• Typically, a single processor
is not powerful enough to run both the simulation and visualization
• Control and, in most cases, rendering will remain local
Types:• Single-processor• Multi-processor• Networked processors
These types can be used in combination
Visualization pipeline can be distributed in a number of ways
Loosely-coupled
Multi-processor - Parallel
Single-processor, possibly multi-processor
Issues
Multi-processor issues• Load balance• Latency• Decomposition, …
Control• Launching remote parts• Interacting with remote parts (steering problem)• Authorization• Authentication• Resource discovery
Data• Format
• Proprietary• Open Standards
• Compression• General purpose• Special purpose
Visual Network Computing/VizserverTM
Multi-processor – loosely coupled Access to SGI high performance
computing/graphics over network• Renders on remote devices• Remote framebuffer compressed
and distributed via TCP/IP over network
• Control over compression Features
• Application transparent• Shared-control• Platform/independent• Advanced visualization
environments• Scalable
Grid
Grid – Development and promotion of standard protocols to enable interoperability and shared infrastructure
• Globus toolkitTM – Open source reference implementation for building grid infrastructure and applications
• Global Grid Forum – Development of standard protocols and APIs for Grid computing
Layered Architecture• Collective – Managing multiple
resources to provide a ubiquitous infrastructure and services
• Resource – Sharing single resources, negotiating access, controlling use
• Connectivity – Talking to things securely
• Fabric – Controlling access and resources locally
Real-time visualization of advanced photon source data, Image from Argonne National Laboratory
Grid Service
Idea: A service with well-defined interface advertises itself in a distributed directory service• Application queries directory
service on how to interact with the service
Web Service• URI• Discovered by XML artifacts• Interactions through XML-based
messages• Standards WSDL, SOAP, …
Grid Service• Extends Web services• Standards – OSGA, OSGI
Grid Visualization
Use Grid Services to discover• Grid Visualization Service• Simulation Running on Grid• Data Stores on Grid
Grid Middleware• Compression• Native / XML Data
Grid Visualization Service• Simulation can register parameters
and data with the service• Data stores or databases can be
registered with the the service• Supports multiple clients• Service manages connections from
external clients• External clients can connect and
interact with data streams• Synchronizes connected clients
Parallel Visualization
Chromium• Open Source• Enables parallel rendering• Replaces systems OpenGL
driver• Industry standard API• Supports existing
applications• Streams of API
• Alters/Discards/Injects• Routes commands• Geometry is moved across
network• Rendered remotely
Visapult - LBNL• Parallel Volume Rendering• Uses OpenRM an industry
standard
Visapult, Image from LBNL
Chromium was created by Greg Humphreys, chromium.sourceforge.net
(a)
(b)
Collaborative Visualization
Radical collocation has proved highly successful• Manhattan Project• Space missions• Software development
Productivity Doubled!• Teasley et al, Michigan
But it requires:• Social disruption• Advance planning• …
Goal of Computer Supported Cooperative Work (CSCW):• Gain in productivity, but reduce
collocation requirement using electronic collaboration
Need to move away from seeing collaborative visualization as a group crowded around a display screen
Towards collaboration over network
CSCW Model
CSCW Model associates applications with approaches
Based on:• When?• Where?
Visualization• Real Time
• Same Place
– AVS, Amira, …
• Different Place
– What do we share?
» Display
» Visualization
» Process
– How many users/location?
Sharing Screen
Simple model• Broadcast display of application
to a set of passive users• Number of available technologies
• IRIS Explorer, AVS, …
• VNC – Virtual Network Computing
– RealVNC – www.realvnc.com
– tightVNC – www.tightvnc.com
VNC, from AT&T
Sharing Visualization
Share the visualization• Geometry is exchanged• Master/Slaves• Number of available applications
• COVISE, IRIS Explorer Advantages
• Greater involvement of collaborators
• Shared Control – Token Passing Disadvantages
• Can’t determine what collaborators are doing
• Limited collaboration
COVISE, from Dr. Ulrich Lang Computing Center University of Stuttgart Visualisation Department
Sharing Process
Each collaborator may participate in producing the visualization
Two variations:• Replicated
• Initial data sharing
• Parameters are interlinked
• Small network traffic
• Application tailored to individuals expertise
• CSVE• Interlinked
• Separate pipelines
• Cross wiring pipelines enables collaboration
• Greater flexibilty
• Varying network traffic
• COVISA
Issues
Portable• Different OS• Different Libraries/Toolkits/MVEs
Functionality• Data• Parameters• Algorithms• Applications
Participation• Joining/Leaving• Floor control• Privacy• WYSIWYTIS• Authentication
System• Performance• Scaling• Reliability• Robust• Security
CSVE, from Patrick O’Leary
COVISA, from Jason Wood, Visualization Scientist, University of Leeds
Access Grid
The Access Grid™ • Ensemble of resources
• Multimedia large-format displays,
• Presentation and interactive environments,
• Interfaces to Grid middleware and to visualization environments
VRVS• Desktop Web-based alternative
Advantages• Greater sense of involvement• Lower geek threshold• Used in combination with VNC
Access Grid, Image from www.accessgrid.org
VRVS, www.vrvs.org
CSVE
Collaborative Scientific Visualization Environment (CSVE)
• Facilitate Scientist - Computer Scientist or Small Group Interaction
• Open Source• Java• JMF• VTK
• Features• Interactive 3D Visualization• Streaming Audio/Video• Streaming Media• Desktop Capture• Chat• Whiteboard• Telepointer*• Remote Control Client*• Data Management*
… a research area expert
A visualization expert interacts with …
CSVE
Anastasia Mironova – Vis 2003 Interactive 3D Visualization
• Handles several data formats• Create/Manage isosurfaces,
slices, …• Simple tools for interacting with
visualization• Seamless network propagation of
visualization parameters
Manage visualization objects …
Create visualization objects …
CSVE
Brian Mullen – Vis 2003 Streaming Media
• Stream any mpeg, avi, mov file to collaborators
Streaming Audio/Video• Stream audio/video from two to
… collaborators Desktop Capture
• WYSIWIS not WYSIWYTIS
Stream scientific videos
Stream audio/video to collaborators
See what they are looking at
CSVE
Scientific Database Currently:
• Relational Database – MySQL, Oracle, …
• Flat files Moving to Meta Catalogue
• Based on an extension of XML Why XML?
• Accepted way of describing things for the Web and the Grid.
• Good at describing things because• Wide range of concepts can be
captured in this way.• It provides a basis for validators,
transformers, parsers, analyzers, displayers, …
• So simple• This is why HTML became so
widely used.• Can teach anyone to use it in a
short period of time.
<?xml version='1.0'?> <list> <recipe> <recipe_name>Chocolate Chip Bars</recipe_name> <author>Carol Schmidt</author> <meal>Dinner <course>Dessert</course> </meal> <ingredients> <item>2/3 C butter</item> <item>2 C brown sugar</item> <item>1 tsp vanilla</item> <item>1 3/4 C unsifted all-purpose flour</item> <item>1 1/2 tsp baking powder</item> <item>1/2 tsp salt</item> <item>3 eggs</item> <item>1/2 C chopped nuts</item> <item>2 cups (12-oz pkg.) semi-sweet choc. chips</item> </ingredients> <directions> Preheat oven to 350 degrees. Melt butter; combine with brown sugar and vanilla in large mixing bowl. Set aside to cool. Combine flour, baking powder, and salt; set aside. Add eggs to cooled sugar mixture; beat well. Stir in reserved dry ingredients, nuts, and chips. Spread in greased 13-by-9-inch pan. Bake for 25 to 30 minutes until golden brown; cool. Cut into squares. </directions> </recipe></list>
CSVE
Message Passing• Objects through bit-stream• Same underlying principles as
remote object broker or RMI
• No parsing
• Flexible
• Extensible
• Efficient
• No parsing!• Moving to XML messages
• The way messages are passed by Grid- and Web-services
• Slower
• Standard format
• Requires parsing messages built into Java
3 Tier Architecture
Application Neuroscience
Pain• Quality of Life• Neurochemical Changes
Image Reconstruction• Removal of Noise and Artifacts• Deconvolution of Light Source• Segmentation of Data
Visualization Techniques:• Maximum Intensity Projection
(MIP)• Volume Visualization
Application Neuroscience
Application Cancer
Bone Cancer• Bone Destruction• Tumor Burdon
Image Reconstruction• Removal of Noise and Artifacts• Edge Detection• Automation• Segmentation of Data
Visualization Techniques:• Isosurfaces• Volume Visualization
CSVE
Interactive Visualization
Desktop CapturePortable•Windows•Apple•Linux
Streaming Media
Additional Applications
CSVE
CSVE
Future Work• Grid Protocol Based
• Resource discovery
– Databases
– Simulations
– Smart Instruments
– Visualization Resources
• Data exchange
• Message passing
• Server as a Grid-service• Remote Control• OpenRM – Direct Volume
Visualization Version• More Visualization Techniques• More sophisticated data
management
Acknowledgements
NSF MRI grant, 0215583, and a NSF REU Supplement to the grant NSF EPSCoR Alaska for funding both Anastasia Mironova’s and Brian
Mullen’s summer research internships The University of Alaska Anchorage (UAA) Office of Undergraduate
Research and Scholarship, Office of Research and Graduate Studies, and Dr. Hilary Davies, whom through Discovery Grants and travel funds made it possible for both Mironova and Mullen to present their work at Visualization 2003
Jonathan Snelling, supported by a NSF REU Supplement, for his work on a multi-document graphical interfaces
Brian Mullen for his development of streaming audio/video tools (he put the “C” in CSVE)
Anastasia Mironova for her development of volume visualization tools, integrating additional data formats, and winning Best Poster at Visualization 2003
My CS 401 Software Engineering class at UAA (Nicholas Armstrong-Crews, Jan Reitspies, Kevin Dickerson, William Sistar, John Vicente, Jeffrey Woods, Daniel Stokley, Justin Dieters, Christopher Johnson, Mark Blum, Shannon Smith, Brandon Douthit-Wood, Shane Ursani, Nathaniel Freeburg, and Christopher Ulmer).
top related