Title: Data, Visualization and Collaboration Research
1Data, Visualization and Collaboration Research
- Jason Leigh, Luc Renambot (UIC-EVL),
- Marcus Thiebaux (ISI),
- Robert Grossman (UIC-LAC),
- Padhraic Smyth (UCI),
- Brian Davis (USGS EDC)
- Donna Cox, Robert Patterson, Michael Welge
(UIUC-NCSA) - Paul Wielinga, Bram Stolk (SARA)
2GOAL
- To bring the network
- end-points into the
- laboratories of scientists
3Year 1 Accomplishments
- Requirements gathering from SIO and NCMIR.
- Developed Straw-person visualization tools for
high resolution montages and volumetric
visualization. - Deployed GeoWall 1 systems to jumpstart
visualization efforts at all domain sites- also
leveraging visualization tools within the Geowall
community. - Defined 1st generation Optiputer Visualization
cluster (32bit) for all sites to replicate. - Experimented with photonic switching to
understand the ramifications of an all-photonic
network on applications. - Designed open web-based standard and
infrastructure for data-mining. - Developed statistical algorithms for data-mining
that are suited for geoscience and
neurobiological applications.
4Year 2 Accomplishments
- Visualization research distributed data access
and performance monitoring. - Collaboration research user studies and
graphics streaming experiments. - Data access and mining research continual
improvement of web services interface and
data-mining algorithms. - Visualization technology deployment and new
development 2nd generation OptIPuter cluster
design based on benchmark evaluations. - Developed the SAGE Scalable Adaptive Graphics
Environment Model
5OptIPuterVisualization Collaboration Research
- SAGE Streaming Graphics for Scalable Tiled
Displays - SAGE Enabled Visualization Tools JuxtaView
Vol-a-Tile - Grid Visualization Utility Point-based
Rendering - TeraVision Display Capture and Streaming
6OptIPuter Scalable Display Systems
UIUC/NCSA
SARA
UCI
USGS EDC
7Scalable Adaptive Graphics Environmentfor
Scalable Display Systems (UIC-EVL)
8GeoWall2 and SAGE Test Trial at SC04
NASA
UIC
5Gb/s utilized total between Chicago, San Diego
and Pittsburgh
9Paul Wielingas Team at SARA(new OptIPuter
partner)
- Have come to similar conclusions on the need for
decoupled rendering and graphics. - Developed complementary pixel streaming
techniques and demonstrated at SC2004. - 6.5Gb/s from Amsterdam to Chicago.
- For iGrid 2005 we will create a persistent
visualization testbed to test collaborative
scientific and biomedical data visualization
between Simon Fraser University (Canada), UIC
(Chicago), SARA/University of Amsterdam
(Amsterdam), KISTI (Korea), CALIT2.
10LambdaCam Web-based Remote Monitoring of
Scalable Displays
11SAGE Future
- Multicasting to support distance collaboration
(11x5 LambdaVision needs 50Gb/s- uncompressed). - New interaction paradigms for manipulating SAGE
windows and controlling applications running
inside SAGE windows. - Social Science Collaboration Team Paul Dourish,
Gloria Mark CALIT2 at UCI
12Multicast
LambdaRAM
JuxtaView
LambdaRAM server
S A G E
Vol-a-Tile
OptiStore
3D points
GVU
GVU Filter
Pre-renderedanimation
Data Mining
Scalable Display (ie 1 to N tiles)
Slow Link
Data Cache Web Service
Laptops
Ethereon
OptiStore2
Varrier
13Multicast
LambdaRAM
JuxtaView
LambdaRAM server
S A G E
Vol-a-Tile
OptiStore
3D points
GVU
GVU Filter
Pre-renderedanimation
Data Mining
Scalable Display (ie 1 to N tiles)
Slow Link
Data Cache Web Service
Laptops
Ethereon
OptiStore2
Varrier
14SAGE Enabled Visualization Tool JuxtaView
(UIC-EVL)
- JuxtaView high resolution image montage display
- Panning and zooming of infinite resolution images.
E.g. USGS Aerial Photographs at 350Kx350K for
120 cities (50TB).
15JuxtaView Also Uses a Networked Memory Cache
System to Hide Latency (UIC-EVL)
- LambdaRAM Network Memory Cache Middleware
- Giant pool of clustered memory to provide low
latency access to large remote data sets by
aggressive use of available network bandwidth. - Relieve application developers from having to
build their own data-prefetching middleware from
scratch.
8-14
all
none
16JuxtaView Visualization Demonstration
Amsterdam(4 nodes)
LambdaRAMServer
Chicago (EVL)(16 nodes)
JuxtaView visualization client
StarLight(4 nodes)
17LambdaRAM LAN, MAN, WAN Performance
3 nodes
4 nodes
15 nodes
- Number of nodes were selected based on resource
availability. - WAN case currently uses 1.5Gb/s out of 3Gb/s.
- Bandwidth utilization increase with more nodes
and latency will drop also.
18SAGE Enabled Visualization ToolsVol-a-Tile
(UIC-EVL)
Orthogonal Views
Navigational Volume
Sub-Volume as voxels
Sub-Volume as isosurfaces
19Grid Visualization Utility(ISI)
- Examine point-based rendering techniques.
- Examine the effect of data partitioning, on
network, and visualization efficiency.
Visualization of signed X-component of the full
motion vector along San Andreas Fault Line
20Grid Visualization Utility(ISI)
- SuperComputing 2004
- Demonstrated multi-site distributed pipeline,
involving clusters at SIO (8 nodes), ISI (8
nodes), and EVL (12 nodes), simultaneously
filtering and feeding data to show-floor PC on
demand, over CAVEWave. - Access 39 GB TeraShake dataset spread across all
nodes.
Used by Southern California Earthquake Center
21 Atmospheric Science Visualization Techniques
Applied to a Volume of Ocean Scientific
Simulation (UIUC-NCSA)
Streamlines colored by vertical velocity. Orange
rising, Blue sinking. Isosurface is colored by
vorticity in the direction of flow.Green is
positive and Purple is negative.
- Single Simulation Time Step Simulation of
stratified fluid flow by Kraig Winters-
computational oceanographer at SIO. - Rendered in Stereo High Definition ie 2 x
1920x1080 - Goal is real time rendering and streaming of
stereo high definition visualizations using SAGE.
22TeraVision Hardware to Capture and Stream
Visualization Signals (UIC-EVL)
- PC High Speed Graphics Capture Card Gigabit
Network Interface. - V3.0 tested at SC 2004 and at recent
demonstration of 10G link to Hawaii - TeraVision broadcast channel interface.
- Multicasting to support collaboration.
- SC04 1280x1024 30fps
- Hawaii 2Kx2K 8bit 30fps
- Now being enabled for High Def Camera
multicasting.
An operator at the National Center for Microscopy
and Imaging Research facility in San Diego, CA
collaboratively navigates the microscope,
communicating with the remote user by Polycom
video teleconferencing
23OptIPuterData Access and Mining Research
- Web services interfaces for high performance
computing - Data mining algorithms using the OptIPuter
24Multicast
LambdaRAM
JuxtaView
LambdaRAM server
S A G E
Vol-a-Tile
OptiStore
3D points
GVU
GVU Filter
Pre-renderedanimation
Data Mining
Scalable Display (ie 1 to N tiles)
Slow Link
Data Cache Web Service
Laptops
Ethereon
OptiStore2
Varrier
25High Performance Web Services(UIC-LAC)
- High performance web services for working with
large and distributed data sets. - XML interface allows users to use a web interface
to do database joins of at high speeds using UDT
reliable UDP-based transport protocol. - Project co-funded by NSF SEIII and MRI and US Army
26OptIPuter Data MinersSpatio-Temporal Data
Mining(UCI and UIUC-NCSA)
- Develop data mining techniques focused on
scientific spatial and temporal data - Extend existing data mining algorithms to online
streaming data using one-look algorithms - One-look algorithms enable real time mining as
data comes in from sensors. - Project co-funded by DOE, NIH, and NSF SEIII
program.
Real time Information Visualization Tools
Cyclone data to track trajectories
27Still Much More In the Works
Personal GeoWall-2
LambdaTable
28Questions?