'Doing Global Earth Science on an Internet Designed for E-Mail'
11.29.2005 -- Fiber optics, according to Calit2 director Larry Smarr, are becoming the key enabler of high-performance science, such as was discussed at last week’s CEOA symposium on the role of universities in developing an integrated, comprehensive, and sustainable global Earth observation system. Smarr was one of the presenters at the symposium.
|
“We’re trying to do global Earth science on the shared Internet, which was designed for e-mail and ftp,” he said. “And we wonder why it doesn’t work!”
Smarr pointed to the cyberinfrastructure for science being provided by the National LambdaRail (NLR) and the TeraGrid at 40x10-Gbps and 4x10-Gbps rates, respectively. Through these activities, some two dozen states around the U.S. now have lambda connectivity. Complementing that is a global set of lambdas at 1-, 2.5- and 10-Gbps rates, the use of which was recently underscored through some 50 real-time demos at iGrid 2005, the first large event held at Calit2@UCSD in late September.
One of the iGrid projects that Smarr highlighted, NEPTUNE, led by John Delaney at the University of Washington, is setting up a “cable observatory” off the coast of Washington state in partnership with NEPTUNE/Canada. (The significance of this project was emphasized by Microsoft CEO Bill Gates who described it at some length in his keynote address at the SC05 [supercomputing] conference in Seattle the week before Thanksgiving.) This project is producing real-time, high-definition video imaging of hydrothermal vents on the seafloor 2.5 kilometers below the sea surface.
“Home high-definition TV,” said Smarr, “needs a 15-megabit-per-second data rate vs. 1,500 megabits for a high-definition camera.” This project recorded full images, compressed them, shipped them over the KU satellite, and provided them in real time to iGrid attendees along with images of Delaney and colleagues on board the research vessel in the area.
|
“NEPTUNE is a prototype of what’s going to be possible,” said Smarr. “Cable observatories enable lighting on the ocean floor, multiple HD cameras shooting stereo in real time, extracting high-res 3-D objects from the imagery, and interacting with biological creatures in the area of the cameras. With instruments capable of 1-Gbps data rates, we’ll be able to have real ‘labs’ on the ocean floor.”
In this context, Calit2 is partnering with the Center for Earth Observations and Applications, which hosted the symposium in the Calit2 building, to look at various scientific areas that could be advanced by the application of new technologies. Calit2 and CEOA partnered in a research exhibit at SC05 and will co-exhibit at the coming American Geophysical Union meeting in San Francisco the week of December 5.
Local environments have to be well instrumented to enable the study of change, said Smarr. Two projects he highlighted are ROADNet and HiSeasNet, prototypes of the future of in situ earth observing systems, the data from which can be assimilated with data collected remotely.
This work is supported by data architectures being developed at the San Diego Supercomputer Center (SDSC) that are integrating sensing devices, data management and workflow systems, and user interfaces to enable users to string components together to perform custom operations.
With sensornets becoming capable of bringing in multiple data types at multiple data rates, the next step is to enable real-time control of those instruments. NEPTUNE, in tandem with a more recent NSF grant called LOOKING, which teams Smarr, Delaney, John Orcutt, and others, will use two-way gigabit streams to/from the sensors and integrate web and grid services to make the management of gigabit data flows “routine.” University researchers need to develop this type of capability because it’s not likely to come from the commercial world.
Putting a national border in the way can even turn out to be a driver for development. The Southern California Coastal Ocean Observing System, described in depth by Eric Terrill from the Scripps Institution of Oceanography in another presentation, provides a rich set of integrated instruments in coastal waters spanning the U.S./Mexico border. “These kinds of cross-border projects are essential for the technology to be adopted globally,” said Smarr, who pointed to CENIC’s and CUDI’s involvement in enabling the link at the San Diego/Tijuana border.
So how does a scientist with a linux cluster in her lab take advantage of this higher-end networking? In answer, Smarr pointed to his OptIPuter project, which is prototyping “campus-level on-ramps to the NLR” to enable user-controlled, end-to-end lambda connectivity right into the scientist’s lab. “Laptop Macs have a natural data flow rate of one billion bits per second,” said Smarr. “That’s higher bandwidth than the bandwidth that connects us over the shared Internet. That shared bandwidth has in fact isolated us from each other. This problem is masked because we can still get e-mail, so we think things are fine. But they’re not. There’s still room for lots of improvement.”
Smarr says that the world is moving to a lambda grid capability. The center of this universe is not the computer, but a 10-Gbps (or higher-rate) optical fabric. Everything else is peripheral. This will allow anyone with lambda access to directly connect a cluster to a data repository located elsewhere to work directly on the data. This architecture will also allow accessing the spare cycles on the lambda-connected clusters. If these capabilities are compatible with the TeraGrid, we’ll have the “Optigrid,” a new term coined by SDSC director Francine Berman, according to Smarr.
The only thing remaining is how to visualize the data. But that’s being addressed too. Smarr described the HIPerWall at Calit2’s UCI division: This system consists of 50 Apple 30-inch cinema displays driven by 25 dual-processor Mac G5s to create 200 megapixels of viewing real estate. Such a tiled wall enables interactive exploration of large Earth science data sets integrated with streaming high-resolution video.
When you put all of these capabilities together, you have a new mode of telepresence, said Smarr. This was demonstrated several months ago via 10-Gbps networks linking San Diego, Seattle, Chicago, and a “loaner connection” to Goddard Space Flight Center in Maryland.
“It enabled an Earth scientist at Goddard to show data locally that was being streamed from San Diego and visualized on the fly,” said Smarr. “This was not trivial to do: It took several dozen of the best network experts in the country six months to set this up. But the effect was amazing! And it’s only a matter of time until this capability becomes mainstream.”
Smarr is convinced we’re at the beginning of a revolution. With the data being generated, it’s now possible to create “global collaboratories.” But for this to move forward, we need applications users that provide the end push, demanding high-quality access.
The iGrid 2005 workshop was a one-week demonstration venue, which was then torn down. Smarr’s goal is to turn the capabilities demonstrated at iGrid into persistent infrastructure.