More than a year ago, Virginia Tech announced the creation of HokieSpeed, a supercomputer so energy efficient that it was the highest-ranked commodity supercomputer in the U.S. on the Green500 List in 2011.
The Green500 compiles a ranking of the most energy-efficient supercomputers in the world, akin to ranking the most fuel-efficient vehicles in the auto industry.
Wu Feng, associate professor of computer science in the College of Engineering at Virginia Tech, designed and built HokieSpeed, with a National Science Foundation instrumentation grant for a modest $1.4 million in the supercomputing financial world. By contrast, Feng’s price tag represented only one-tenth of a percent of the cost of the 2011 Top500’s No. 1 supercomputer, the K Computer from Japan.
Feng believes he and his team can do even better. So does the Air Force’s Office of Scientific Research (AFOSR).
The AFOSR has awarded Feng, formerly of Los Alamos National Laboratory – where he designed Green Destiny, a highly energy-efficient, 240-node supercomputer that consumed 3.2 kilowatts of power (equivalent to two hairdryers) and occupied five square feet in 2004 – a core award of $3.5 million over the first three years of the contract with the option for a two-year extension at another $2.5 million, or a total of $6 million over five years to pursue a large-scale interdisciplinary effort.
The goal of the Air Force contract is to achieve a substantial increase in the simulation speed of computational fluid dynamics of its micro air vehicles (MAVs), a class of unmanned aerial vehicles, using accelerator-based supercomputers like HokieSpeed.
To achieve this target, Feng said his objective is to herald a new age in multi- and many-core parallel computing that he believes will “transform supercomputing.”
The key to Feng’s approach is to couple innovative advances in algorithms and mathematics with engineering progress in the co-design of hardware and software in supercomputing.
He has pulled together an internationally recognized team of researchers from Virginia Tech and North Carolina State University into the effort. They represent aerospace and mechanical engineers, in addition to his home department of computer science, as well as mathematicians.
From Virginia Tech, the engineering members of the interdisciplinary team include: Christopher Roy, associate professor of aerospace and ocean engineering; Adrian Sandu, professor of computer science; and Danesh Tafti, professor of mechanical engineering. Eric de Sturler , professor of mathematics in the College of Science at Virginia Tech, is also a member of the team.
In computing, Moore’s Law observes that the number of transistors on integrated circuits doubles approximately every two years. This prediction made in 1965 by Intel co-founder Gordon E. Moore, has proved accurate for almost half a century and guides long-term planning in the semiconductor industry and has correlated to substantial performance gains every two years as well.
“But much of the realizable performance remains untapped” as many domain scientists and engineers are still learning how to “fully exploit parallel hardware and co-design software for performance gains,” Feng explained. “Furthermore, coupling hardware-software co-design with advances in algorithmic innovation offers the promise of multiplicative speed-ups.”
Feng has already achieved significant speed-up in recent supercomputing research, leveraging a co-design approach with hardware, software, and algorithms for problems in bioinformatics, biophysics, neuroscience, and seismology.
Feng’s new interdisciplinary team plans to develop computational fluid dynamic (CFD) codes and a supporting hardware-software ecosystem for the simulation of micro air vehicles (MAVs). Based on his earlier work, Feng said the Virginia Tech research team should be able to “achieve substantial speed-up over current simulations and provide significantly better utilization of the underlying and co-designed hardware-software of a supercomputer.”
The micro air vehicle (MAV) can be as small as about five inches with an aircraft close to insect size expected in the near future. In addition to having numerous military applications, these tiny robotic instruments are useful in hazardous conditions such as collapsed buildings or in nuclear power plants.
The team’s efforts should support the aerodynamic predictions for these increasingly tiny vehicles.