AsianScientist (Aug. 15, 2018) – When something goes wildly viral—Beyoncé’s triumphant comeback performance at Coachella, for example, or videos of oversized cats desperate to fit into tiny boxes—we often say that it ‘broke the internet.’ Yet, that dubious honor might well belong to the data-hungry scientific community, which on a daily basis generates and moves orders of magnitude more information across networks than Queen Bey and YouTube cat Maru’s traffic combined.
For information to flow at that scale, biologists sequencing genomes, high-energy physicists tracking subatomic particles and astronomers training giant telescopes at the heavens need something more than the World Wide Web, said Dr. Joe Mambretti, director of the International Center for Advanced Internet Research (iCAIR) at Northwestern University in the US.
“The general internet cannot handle [such huge amounts of data]—it’s not designed for it. In fact, the internet actually has blockages in it to prevent big flows of data, because that could mean a denial of service attack,” he told Supercomputing Asia.
If the internet is a congested city street, Mambretti and his collaborators are assembling a six-lane highway for information: a sprawling computing architecture called the Global Research Platform (GRP), which will help researchers around the world access and exchange large datasets.
“They need special channels and special capabilities for moving this data,” explained Mambretti.
Connecting the dots
The GRP has its roots in the Pacific Research Platform (PRP), a regional data-sharing system connecting research universities and supercomputing centers on the US West Coast, as well as several partners elsewhere in the US and in Asia.
The PRP consists of multiple science ‘demilitarized zones’(DMZs)—a model developed by the US Department of Energy in which a secure, high-performance network for data-intensive science is kept separate from the general internet.
But while science DMZs already existed at multiple individual research campuses across the country, it wasn’t until the PRP—funded by the US National Science Foundation in 2015—that these islands were linked up into a unified network, said Dr. Thomas DeFanti, one of the PRP’s co-principal investigators and a research scientist at the University of California, San Diego’s Qualcomm Institute.
To handle the huge flows of data coming through the network, DeFanti and his colleagues designed low-cost networked PCs called FIONAs (an acronym for Flash Input/Output Network Appliances), which today hum at many of the PRP’s nodes. Equipped with flash memory and powerful GPUs, FIONAs send, receive, monitor and cache data, acting as a data ‘capacitor’ or interface between the high-speed network and the node’s scientific instruments and computers.
In some cases, the PRP has sped network connections up by a thousand-fold over the general internet, said DeFanti, who is also distinguished professor emeritus in computer science at the University of Illinois.
“We’re very proud of that… but the question is how do we do this in more than just California and a few other places?”
Going global
Mambretti, DeFanti and their collaborators now want to work with partners around the world to replicate the PRP on a global scale. The GRP team holds regular demonstrations and workshops to get the word out, including at major high-performance computing and networking events such as the Supercomputing 2017 conference and the Global Lambda Integrated Facility’s Global LambdaGrid Workshop 2017.
The field is starting to move away from a one-size-fits-all approach to networking for data-intensive science, acknowledging that each scientific community has its own specialized needs and hence requires specialized capabilities, said Mambretti. The GRP’s architecture will address this by allowing networks to be segmented—that is, sliced up to provide such differentiated services, he added.
Indeed, various scientific communities have already expressed interest in the new platform. The GRP team is planning to work with the astrophysics community on data from the Square Kilometer Array, the Large Synoptic Survey Telescope and the Nobel Prize-winning Laser Interferometer Gravitational-Wave Observatory; with the high energy physics community at facilities such as the Large Hadron Collider and Fermilab; and with precision medicine initiatives in countries such as the US, the UK and Taiwan.
Research institutions and supercomputing centers in Asia—in China, Japan, Singapore, South Korea and Taiwan, for example—have also registered interest in being part of the network, said Mambretti.
Getting partners on board, however, is not without its challenges.The GRP’s novel architecture has proved a stumbling block for more conservative organizations, said Mambretti.
“They have their traditional tools, traditional ways of doing things and traditional policies, and they’re suspicious of the new,” he explained. “It’s easier for [networking] to be one-size-fits-all… when people want different things, it’s harder for you as a provider.”
Time, perhaps, will convince doubters of the worth of the new platform. While the GRP is not yet fully up and running, Mambretti and DeFanti stressed that all the necessary tools, as well as many collaborations, are already in place for the network to take off.
“We already have all the LEGO blocks—we just need to put them together,” said Mambretti.
This article was first published in the print version of Supercomputing Asia, July 2018.
Click here to subscribe to Asian Scientist Magazine in print.
———
Copyright: Asian Scientist Magazine; Photo: Shutterstock.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
