Increasing the Reliability of Supercomputing

The rate of progress in com­puting and tech­nology is increasing at warp speed. A commonly- held com­pu­ta­tional truism, known as Moore’s Law, asserts that com­puter processor per­for­mance dou­bles every 18 months. But some of the basic sup­port sys­tems are strug­gling to keep up, and it’s becoming increas­ingly chal­lenging to main­tain the com­puting processor’s reli­a­bility and effi­ciency, which leads to high oper­a­tional cost

For instance, Titan—the largest super­com­puter in the U.S., and the com­puter behind some of the most impor­tant sci­en­tific research in the world—requires eight megawatts of energy to run at peak per­for­mance. That’s enough elec­tricity to power a small city.

At roughly $1 mil­lion in energy costs per megawatt, run­ning the com­puter takes nearly $9 mil­lion per year alone, without accounting for any of the other costs asso­ci­ated with run­ning the world’s third- largest supercomputer.

Devesh Tiwari, a newly appointed assis­tant pro­fessor of elec­trical and com­puter engi­neering in the Col­lege of Engi­neering, says this just isn’t sustainable.

Tiwari is familiar with Titan and its oper­a­tional needs. Prior to joining North­eastern this semester, he worked as a staff sci­en­tist at the Oak Ridge National Lab­o­ra­tory, which is funded by the U.S. Depart­ment of Energy and where the super­com­puter is housed.

There, he was instru­mental in designing and imple­menting inno­v­a­tive tech­niques that improved the resilience and effi­ciency of the high- powered com­puting system, which enhanced the sci­en­tific pro­duc­tivity of its users.

“When you have a sci­en­tist studying things like astro­physics or neu­rology, the com­pu­ta­tional part can get in the way of his true research if it’s not working cor­rectly,” Tiwari said. “The tech­nology has to be really solid and really dependable.”

Tiwari’s research focuses on just that: devel­oping new ways to make large- scale com­puting sys­tems sus­tain­able by making them more resilient and more energy- efficient, thereby reducing oper­a­tional costs and increasing the pro­duc­tivity of the sci­en­tists using the computers.

His work so far has not only helped sta­bi­lize the country’s largest super­com­puter, but has been rec­og­nized in highly com­pet­i­tive inter­na­tional sci­en­tific jour­nals as well.

And it’s not just far- flung super­com­puters that need to be updated. In a world that increas­ingly relies on vast com­puter net­works for basic func­tions like banking and elec­tricity, much of these sys­tems are alarm­ingly “ad- hoc and sub­op­timal,” Tiwari said.

“My big dream—which I believe is achievable—is cre­ating a society where the com­plex infra­struc­tures of health care, med­ical devices, advanced man­u­fac­turing, trans­porta­tion sys­tems, and power grids are sus­tain­able and resilient,” he said. “It has to be healthy all the time. Only then can we truly build on it to make progress.”

Thus the fierce race to build the world’s most pow­erful super­com­puter mir­rors the drive to build a more sus­tain­able, robust society; it all depends on ensuring the infra­struc­ture behind it all is up to the task.

So Tiwari and his wife—both having grown up in India’s rel­a­tively warm climate—overcame their aver­sion to the cold so they could come to Northeastern.

Here, Tiwari found the “dynamic fac­ulty,” “inno­v­a­tive spirit,” and “encour­aging, exploratory envi­ron­ment” that matched his own pas­sion for his field.

“We had never imag­ined that we would even con­sider moving to a city with rel­a­tively intense, cold win­ters,” he said. “But, Northeastern’s ener­getic envi­ron­ment and encour­aging cul­ture of inno­va­tion made us change our minds. We are delighted to be part of the North­eastern family. As my father says, ‘One of the best ways to give back to society is through con­tin­uous inno­va­tion and shaping the minds of younger gen­er­a­tions through teaching.’ We are lucky to have found a place where we can do just that and call it home.”