The exponential growth of scientific data has put considerable strain on existing research networks. Just five years ago, the United Nations International Panel on Climate Change (IPCC) was fed with 35 TB of data to use in developing its assessment report. In two years, when the next IPCC report is published, it is estimated that dataset will be up to 2 petabytes, more than a 57-fold increase.
To support the transfer of these massive buckets of bytes and advance adoption of faster communications technology, the government introduced the Advanced Networking Initiative (ANI). The program that was created with a $62 million pot of money that was scooped out of the 2009 federal stimulus package, also known as the American Recovery and Reinvestment Act (ARRA). At the heart of the initiative is a prototype 100Gbps testbed network, built by the Energy Sciences Network (ESnet), in collaboration with the Internet2 consortium.
The network currently connects the National Energy Research Computing Center, (NERSC) in California, the Argonne Leadership Computing Facility (ALCF) in Illinois, and the Oak Ridge Leadership Computing Facility (OLCF) in Tennessee. Brian Tierney, head of ESnet’s Advanced Networking Group, mentioned the network’s popularity with scientists in a recent article on the organization’s website.
“Our 100G testbed has been about 80 percent booked since it became available in January, which just goes to show that there are a lot of researchers hungry for a resource like this,” says Tierny.
While bandwidth is an integral component to moving information between facilities, applications also determine how effectively that data is transferred. The Climate 100 collaboration, also born from the ARRA, was tasked with developing new methods for moving extremely large amounts of climate data.
Mehmet Balman of the Berkeley Lab’s Scientific Data Management group and member of the Climate 100 collaboration, explains that advanced middleware applications are needed to handle the variety of small and large data across high-throughput networks. The Climate 100 group used the 100Gbps network as a testing environment for their applications.
The Climate 100 tool was used on the ANI testbed to demonstrate a 35 terabyte transfer of data between NERSC and ALCF. The operation took roughly 30 minutes to complete. Compared with a 10Gbps, the same transfer would have taken roughly five hours.
The ANI project is set to wind down in a few months, after which, the test network will be folded into ESnet’s fifth-generation production infrastructure.
Long distance networking has become a familiar bottleneck in scientific computing. As the dataset sizes continue their upward climb, these resources will be taxed even further. Projects like ANI display forward thinking from the government, at least when the federal money is flowing, and demonstrates the enabling effects of 100G bandwidth.