Wu-chun Feng
The Wizard of Los Alamos
December 2004 / January 2005
Volume 2 Number 6
By holding more patents by far than anybody else, and by creating such things as the stock ticker, the multiplex telegraph, a voice recorder and, of course, the light bulb, Thomas Edison was known as the "Wizard of Menlo Park," the small town in New Jersey where he did most of his inventing.
Edison, by the way, aided the U.S. in developing some 40 inventions dealing with anti-submarine warfare after German U-boats sunk the Lusitania in 1915. And at about that time he also demonstrated his prescience by telling The New York Times that "the government should maintain a great research laboratory" and proposed collaboration between inventors and national defense.
At one of those "great research laboratories"—€”Los Alamos—€” Wu-chun Feng is following in the Edison tradition. Not only does Feng work on the nation's defense, but he is also one of the great inventors of the computer age. Among Feng's inventions are three that recently won R&D 100 Awards, given annually to the 100 most influential inventions of the previous year. Feng's inventions are the Green Destiny energy-efficient supercomputer, the 10-gigabit Ethernet adapter and the computational biology program, mpiBLAST.
Dr. Feng's Green Destiny represents a new approach to supercomputer design. Instead of being obsessed solely with clock speed, Green Destiny trades off clock speed for high reliability. The result is an efficient supercomputer that does not require expensive cooling systems and almost never breaks down (component failure is a major problem with supercomputers). Because it can run continuously, Green Destiny performs more computations per year than supercomputers with many times Green Destiny's clock speed. Last month, Orion Multisystems began selling desktop supercomputers based on Feng's designs for under $10,000. Supercomputers typically come with pricetags that sometimes exceed several million dollars, so Green Destiny-based supercomputers are poised to bring supercomputers to medium and small companies and, eventually, to the home.
Last year Feng worked with a team that smashed the Internet2 Land Speed Record. The team included scientists from Lawrence Livermore National Laboratory, Caltech, the European Center for Nuclear Research and the Stanford Linear Accelerator Center. A key component of the record-breaking effort was the 10-gigabit Intel Ethernet adapter optimized by Feng and his team. Ethernet handles over 95 percent of all internet traffic. Feng's Ethernet adapter, even when limited by a slower trans-Atlantic network link, was able to transmit a terabyte of information in less than an hour from California to Switzerland. A terabyte is a staggeringly large amount of data (1,500 CDs). Using the Ethernet adapter, an entire CD can be downloaded in less than three seconds. By besting the previous internet speed record by two and a half times, Feng's work not only won an R&D 100 Award but also earned a listing in the Guinness Book of World Records.
Feng's third R&D 100 winner is mpiBLAST, a parallel version of BLAST. Currently, BLAST is the most-run piece of software in all of molecular biology. As much as 95 percent of all computing time at pharmaceutical and biotech companies is dedicated to BLAST searches. BLAST allows molecular biologists to compare new genes to all known genes. For example, in the human genome project (which also began at Los Alamos), BLAST can compare a newly isolated human gene to all known mouse, bacteria, yeast and frut fly genes. BLAST can discover if the new gene likely plays a role in muscle movement, in immune systems or in brain function. Similarly, BLAST identifies emerging viruses such as SARS so they can be tracked and treated effectively.
The largest problem prior to Feng's contribution was BLAST's extremely slow speed. A single BLAST search could take about 24 hours. Feng invented an improved, parallelized version of the program called mpiBLAST. Instead of taking a full day and night, an mpiBLAST search can be completed in less than seven minutes. Naturally this has quickened biopharmaceutical research enormously.
Unlike other inventions of his, Feng did not patent mpiBLAST and has no financial interest in its success. Instead, he released it as open-source code on the internet—€”making mpiBLAST free to all users. "Given the importance of this tool to bioinformaticists and, hence, society as whole, I figured that making it free was the least that I could do," says Feng.
Life of the Serial Inventor
Feng's first significant invention was a "driving directions" program he wrote back in the 1980s when he was an undergraduate at Pennsylvania State University. He invented the program while sitting in an uncomfortably air-conditioned computer room over Christmas break. He typed code wearing "a down coat and knit gloves." Of his first inventive experience, Feng says, "There were a lot of lonely hours. I was using this cryptic language to access directions from point —€˜A' to —€˜B.' Who in the world would want to use such a language? What I thought at the time was well, gee, some type of natural language interface and processing [would] allow plain English text to be turned into code." Instead of requiring lines of gibberish-appearing characters, Feng's new program let people ask for directions using normal English. Although others independently invented it, this type of natural language interface is now widely used in map searches and in luxury vehicles. Of his first invention, Feng says "For me it was just kind of a challenging problem. Now looking back, I really had an idea that would end up in a car. It actually came to fruition."
As a graduate student at the University of Illinois, Feng was well-placed to make a major impact on computer networking. Mosaic, the first Web browser, had been developed there. Netscape founder Marc Andreessen worked in the same lab as Feng's wife. At the university, Feng worked on computer protocols that allowed for streaming video a year before Progressive Networks (later RealNetworks) announced RealVideo. "For two years," says Feng, "we broadcast the Indy 500 over the internet."
Interspersed among his multiple degrees, Feng worked for IBM's elite T.J. Watson Research Center, for NASA and for streaming-video Vosaic Corporation. He also held adjunct faculty positions at Purdue University, the University of Illinois and Ohio State. Now at Los Alamos, Feng leads the Research and Development in Advanced Network Technology (RADIANT) team in the Computer and Computational Sciences Division.
Earlier this year the Chinese Institute of Engineers named Feng the "Asian American Engineer of the Year." Feng credits his inventive success in large part to his heritage. "For Asian immigrants, particularly Chinese, the bar for —€˜acceptable' performance is not a —€˜C,' it's an —€˜A.'" According to Feng, the Asian-American immigrant culture also keeps him from resting on his laurels. A typical response to success, says Feng, is "—€˜OK, that's good. What's next?' Nothing you can do is good enough."
Keeping track of inventions is a challenge for any prolific inventor. When Thomas Edison died, he left behind almost five million pages of scratch paper filled with new ideas. Feng has faced a similar problem with mixed success. At first, says Feng, "I wrote on my hands a lot." While that worked fairly well ("I see it, I have to do something with it as soon as possible"), he was somewhat embarrassed by having ink-stained hands. Currently Feng's pockets are filled with notes jotted on 3 x 5 cards and hotel stationery. Writing inventions down on scraps of paper is impossible, however, when Feng is out running or riding his bike. In these cases, he has to remember the inventions, and this is also problematic. "I will admit," says Feng, "that my maximum —€˜cache capacity' for ideas is eight." When riding, Feng takes the eight ideas, turns them into key words and makes an acronym out of them so he can remember the ideas when he gets back to the office. "I sometimes get lazy and call in and leave voice mail for myself from my cell phone."
At the end of our interview, Feng started talking excitedly about inventing a way to keep track of his ideas with the interactions between ideas correlated and mapped out. Perhaps this will be the next invention.
Jeffrey Stewart is a business development executive at LANL's Technology Transfer Division.
Copyright © 2012 | Innovation America