BIG DATA

In the 2007 article, “The End of Theory," Wired editor-in-chief, Chris Anderson, predicted that in the near future, computers would cull immense databases of observational measurements to produce new insights about the natural world. This data deluge, he argued, would make the scientific method obsolete.

“Instead of coming up with theorems and running the simulations to see if they actually fit the observations, you take very large amounts of data collected from the real world and perform analyses on that data to derive the mathematics behind it, or apply statistical methods to attempt to understand various phenomenon,” explained Chris Jordan, senior operating systems specialist, responsible for data infrastructure at the Texas Advanced Computing Center (TACC).

This kind of discovery requires a central repository for massive data collections — satellite scans of Earth, digital negatives of plants — that can be accessed and manipulated, compared and reviewed, easily, and from any location.

TACC’s newest resource, Corral, which debuted on March 31, 2009, is just that.

With 16 Dell server nodes and 1.2 petabytes of DataDirect Networks storage — twice the space required to hold the entire Netflix DVD catalogue, and four times larger than any current data-collection resource on the TeraGrid — Corral will effortlessly handle the challenges and opportunities of data-driven science.

“We’re ahead of the curve in terms of providing this kind of dedicated data collection and application resource,” Jordan said. “A lot of other sites are doing data collections, but very few sites are providing this kind of universally accessible, unified resource.”

Besides the drive for new, data-driven scientific methods, there were several other reasons to build Corral, says Jordan. First, larger and more sophisticated computing and visualizations systems, like Ranger, Lonestar and Stallion, create tremendous amounts of data that needs to be properly stored and managed. Second, traditional data-collection repositories, such as museums and physical archives, are being revamped and re-imagined for the 21st century. Corral addresses the need for the digital preservation of important documents and specimens, while allowing archives to share their materials more broadly than ever before.

With data-transfer capabilities approaching the blazing rate of six gigabytes per second, Corral adds a dynamic new element to TACC’s high-performance computing arsenal. Corral's built-in services and software tools allow researchers to derive the greatest insights from their data, and the system’s direct connectivity to TACC’s other computational resources, including Ranger(one of the most powerful supercomputers in the world), and Stallion (the world’s largest tiled visualization display), limits time wasted in data transfer.

Since Corral went online in March, six diverse projects, from archeology to engineering, have begun using its capabilities to extend, and even generate, new scientific applications.

Alaskan Archives — Plants, Whale Songs and Dinosaur Fossils

One of those projects, the Alaska Herbarium (ALA), developed by the University of Alaska’s Museum of the North, preserves and circulates plant specimens from the region. Through a grant from the National Science Foundation, the museum is in the process of digitizing all of their specimens.

“As you may be aware, Alaska is fairly difficult to get to,” said Steffi Ickert-Bond, assistant professor of Botany and curator of the University of Alaska Museum of the North Herbarium. “Even within Alaska it’s quite hard for rural communities to come to Fairbanks. In our effort to digitize the specimens and make them available over the Internet, we hope to engage researchers from all over the world to see what we have in the herbarium.”

230,000 specimens of Arctic bio-diversity have been diligently collected over the decades. However, until recently, the collection was confined to the museum’s home archive.

With over 1.2 petabyte of storage, data-transfer speeds up to 6 gigabytes a second, and direct connectivity to TACC’s other computational resources, Corral is built to handle data-intensive problems.

“Data storage had been a real problem for us,” Bond admitted. “When we started this project, an open web repository of biological images called MorphBank was going to support all of our images at the Florida State University in Tallahassee. But they weren’t prepared for the masses of images that we were producing. It became apparent very quickly that they couldn’t handle our storage and web-accessibility requirements.”

The specimen files — high-resolution photo negatives required for fine-grained comparison — were large, and the museum needed not just a site to store the data, but a way to make the images available to the scientific community.

Enter Corral, the only system currently capable of doing the job. “We came up with a solution through the TeraGrid, of which TACC is a partner, and they were just incredible,” said Bond. “We’re now able to serve all of our images online, both in digital negative format and as jpegs, with tape backup storage at the San Diego Supercomputing Center. Within seconds of taking an image, we’re able to store the image and make it available for viewing by the public. It’s been a fruitful collaboration.”

As a unique plant record spanning decades, the herbarium’s collection serves as an important benchmark for global climate change, drawn upon by researchers around the world. “We’re already experiencing some of these changes,” Bond said. “There is an up-migration of shrubbery that is effecting native flora; invasive plants are spreading. Because of climate change, we have had an influx of certain pests that are affecting our native flora. It’s happening all over, and we’re in danger of not even knowing what we’ve lost.”

Only six months into the collaboration, all of the herbarium’s digitized materials are archived on Corral and available to researchers via the Arctos portal, which pulls the files across TACC’s high-bandwidth network.

Oxytropis vassilczenkoi subsp. substepposa Jurtzev, Western Beringian endemic legume (Fabaceae) collected in Russia, Western Chukotka, Mount Nagleynyn, SW shore of Chaun Bay, Vetrechnyi Creek (4-5 km south of the mountain), 06 Jul 1968 (link to high-resolution version) 

In addition to its extensive Alaskan plant collection, the Herbarium has the largest collection of Russian specimens outside of Russia.

Getting its large archives to a secure and stable location is an important first step for the herbarium. Ultimately, using database software currently available on Corral, researchers will be able to do complex data-driven analysis based on the archival files, for instance comparing specimens from different geographic and temporal locations to map how climate change is affecting the Arctic.

Based on the successful relationship between TACC and the University of Alaska, Corral is also hosting a large collection of killer whale songs, and is developing a partnership with University of Alaska Museum of the North to digitize and host the specimens from a database of Arctic dinosaur fossils.

Bigger Deluges of Data

At approximately five terabytes, the University of Alaska’s digital datasets are some of the smaller collections on Corral. The big players, whose funding commitments made Corral possible, are The Center for Predictive Engineering and Computational Sciences (PECOS) project from the Institute for Computational Engineering and Sciences (ICES) at The University of Texas at Austin, and research at the Center for Space Research, also at The University of Texas at Austin [see descriptions below].

These projects anticipate using more than 100 terabytes of storage each and are greatly helped by the fact that Corral is directly linked to TACC’s large computational and visualization systems.

“You can bring your data into Corral and do some computation using Ranger, perform some further analysis using Spur, and then visualize the data at the Vislab, without ever moving any of the data around,” Jordan said.

TACC expects that within two years, Corral, the largest of its kind at a supercomputing center, will be completely full. For that reason, the center designed the system so it could be doubled, tripled or made ten times larger.

With its suite of high-performance and data-intensive systems in place, Jordan believes TACC is well positioned to be at the forefront of advances in data-driven applications and analysis.

“The advantage of having Corral is that we have the ability to offer services based on new methodologies, and to support them in a very flexible way,” said Jordan. “This gives us the opportunity to learn what some of the best practices are and share that information with a wide variety of projects.”

Steffi Ickert-Bond's work is supported by grants from the National Science Foundation Division of Biological Infrastructure. 

Aaron Dubrow and Faith Singer-Villalobos
Texas Advanced Computing Center
Science and Technology Writer