110 likes | 136 Views
Explore how UMass/CASA utilizes experimental weather radars to generate accurate nowcasts for hazardous weather detection. The demonstration showcases a dynamic end-to-end nowcasting system on GENI, utilizing shared sensing, networking, storage, and computing resources on-demand.
E N D
Nowcasting: UMass/CASA Weather Radar DemonstrationDavid Irwin November 3, 2010 http://geni.cs.umass.edu/vise http://geni.cs.umass.edu/dicloud http://www.geni.net
Problem • CASA (an NSF ERC) is studying experimental networks of small controllable weather radars • Better data is the foundation of better hazardous weather detection and earlier warnings • Complex modeling to detect inclement weather requires many resources: sensors, bandwidth, storage, and computation • Costly to dedicate resources for rare events • How do we generate accurate, short-term “nowcasts” using these new distributed radar systems?
Solution • Today: only a few large NEXRAD radars (100s) • Tomorrow: many (1000s) smaller, less expensive radars produce data close to the ground where weather happens • Requires a flexible infrastructure for coordinated provisioning of shared sensing, networking, storage, and computing resources on-demand
Example: Puerto Rico Testbed • UPRM Student Testbed • Led by Jorge Trabal, Prof. Sandra Cruz-Pol, and Prof. Jose Colom • http://www.youtube.com/watch?v=7TR64BhwMlI
Demo Background • Dynamic end-to-end Nowcasting on GENI • Use GENI/Orca Control Framework (RENCI/Duke) • https://geni-orca.renci.org/trac/ • http://geni-ben.renci.org:11080/orca/ • Reserve heterogeneous slice of resources • Sensing Slice: UMass ViSE radars • Networking Slice: NLR, BEN-RENCI • Computation Slice: Amazon EC2 + UMass and RENCI VMs • Storage Slice: Amazon S3
Demo Data Flow • Dynamic end-to-end Nowcasting • Mapping Nowcast Workflows onto GENI archived netcdf data aggregated multi-radar data Nowcast images for display “raw” live data Radar Nodes Archival Storage Upstream LDM feed Nowcast Processing Post to Web
Generate “raw” live data ViSE/CASA radar nodes http://stb.ece.uprm.edu/current.jsp • Ingest mulit-radar data feeds • Merge and grid multi-radar data • Generate 1min, 5min, and 10min Nowcasts • Send results over NLR to Umass • Repeat Use proxy to track usage-based spending on Amazon and enforce quotas and limits http://geni.cs.umass.edu/vise/dicloud.php ViSE views steerable radars as shared, virtualized resources http://geni.cs.umass.edu/vise “raw” live data Nowcast images for display • DiCloud Archival Service (S3) • LDM Data Feed (EC2) Multi-radar NetCDF Data Nowcast Processing
GENI Technologies and Credits • UMass-Amherst • ViSE and DiCloud projects • University of Puerto Rico, Mayaguez • Jorge Trabal, Prof. Cruz-Pol, and Prof. Colom • OTG Radars • Colorado State University • Prof. V. Chandrasekar • Nowcasting Software • RENCI/Duke • Orca Control Framework • BEN network • Starlight
Conclusion • GENI is critical for next-generation applications • Enable nowcasting in experimental radar systems • GENI capabilities: “sliceability”/virtualization, federation, network programmability • Provide domain scientists a new platform • Experiment with tightly integrated systems combining sensing, storage, networking, computing • Engage domain scientists in CASA and elsewhere • Extend GENI network to Puerto Rico