Li-Shiuan Peh puts power savvy into network design
By Sara Peters

Editor’s Note: War in the Middle East, dwindling fossil fuel supplies, and the increasingly high-tech lives of everyday people are bringing energy issues to the forefront of public discussion.

This series will highlight some of the energy-related research being performed by faculty in the School of Engineering and Applied Science.

Photo by Frank Wojciechowski Li-Shiuan Peh, assistant professor of electrical engineering, conducts research on power-aware networks.

We face our daily lives armed with laptop computers, cellular phones, personal digital assistants, and an arsenal of other high-performance technical devices in our offices, homes, cars, and backpacks.

These devices help us survive our high-speed existence by performing extraordinary, Jetsons-like feats … until their batteries run out or their circuits overload.

The power hunger of high-tech tools is ravenous, and in a world getting short on energy sources, the need for more power efficiency is considerable.

Luckily, the research group of Li-Shiuan Peh, assistant professor of electrical engineering (EE), is making breakthroughs in the creation of power-aware networks. Power-aware networks are an element essential to forging the technology of the future.

Professor Peh’s group attacks the network power problem on three main fronts.

First, through conferences, papers, and informal discussions, she and her group push industry and academia to increase the power awareness in networks.

Second, they developed tools for estimating network power use over time, detecting peaks and troughs in energy consumption.

Third, they are now focusing their attention on designing a power-aware network.

Network fun
Professor Peh wants to be clear that she does not work on the Internet, but, rather, networks within an individual computer. She clearly enjoys this work, beginning or ending most of her statements with the phrase, “It’s fun.”

“ On the Internet there are a lot of regulations, but not on these networks,” she said. “When you design networks like this, you can decide exactly how you want to connect things. You can design your own roads. You can design your own rules. Everything is kind of up for grabs.”

This freedom results from the fact that these networks have mostly been proprietary designs, with no clear standards emerging yet.

“ While power-aware networks can benefit networked computers in commercial data centers, there hasn’t been an immediate need for power-aware networks in consumer products because most of our laptops and gadgets still operate with a single processor,” Professor Peh said. “But more and more, applications are becoming a lot more power-hungry, and so almost all of industry is planning on moving up to multiple processors. IBM is starting to have four. Intel says they’re looking at eight. So when you have many of these working together, they need a network to communicate through.”

Power has not historically been a matter of much concern to network designers. Professor Peh says that performance has always driven the innovation, with little or no thought to power.

Yet although she believed power should be addressed, the question remained: How significant a limitation is energy to a network? No one had the answer.

Energy estimate
Thus, ORION was created. ORION is a performance simulator used to estimate network power use over time. ORION stands for Open Research Infrastructure for Optimizing Networks, but was mainly named after the brightest constellation—or network of stars—in the sky.

“ The naming is the most fun part,” Professor Peh said.

Professor Peh and EE Professor Sharad Malik lead the ORION project.

EE graduate students Hangsheng Wang, Li Shang, Xinping Zhu, Xuning Chen, Noel Eisley, Vassos Soteriou, Amit Kumar, Julia Chen, Kevin Ko, and Eric Chi and computer science graduate student, Yong Wang round out the ORION team.

Professor Peh explained that prior methods of estimating network power simply counted the number of “hops” a piece of information had to travel and multiply it by the average power used per hop. ORION’s method looks at the actual power use, not the hops, and detects and accounts for peaks and troughs in power use.

“ Our argument is that you have to look at power itself,” Professor Peh said. “Let’s say you’re concerned about battery life. How long a battery lasts depends on how it is used over time; how high of a peak it can hit. It’s not just about how many joules it uses.

“ Or let’s say you’re looking at a laptop and thinking about temperature. Whether or not this thing burns up depends on how hard it’s working at one moment. That’s why we said you need this temporal understanding.”

Researchers at Intel, and faculty at Stanford, Wisconsin, and Illinois universities are using ORION in various applications, including the design of on-chip networks and cache connection.

Network design
Use of ORION has confirmed the need for energy-aware networks.

“ But while we’ve been saying this,” Professor Peh said, “we haven’t really demonstrated a prototype. So that’s the angle we’re working at now—actually designing this power-aware network.”

Her group is working with EE Professor Paul Pruçnal and Harvard University Engineering Assistant Professor Gu-Yeon Wei.

“ Gu-Yeon does electrical circuits, Paul does optical circuits, and I’m the router guy,” Professor Peh said.

One element of creating a power-aware network is addressing the high temperatures generated by a hard-working network.

Enormous fans to cool a steamy network down are expensive and not very scalable. So, Professor Peh’s group, with EE Professor Niraj Jha, has come up with something else—a temperature-management technique called Thermalherd.

“ It’s like a shepherd of temperature. I told you the naming was the most fun,” Professor Peh said. “Thermalherd senses where there is a jam, and tries to reroute the traffic away like a policeman. It eases the hotspots.”

Peh’s group is combining their collective skills to examine every aspect of networks to unearth other solutions.

“ That’s one thing I sell to my students,” Professor Peh said. “We don’t just look at one thing. We look at everything—theory, algorithms, circuitry, architecture.”

Up next
Looking forward, Professor Peh would like to study the increasing power-efficient possibilities of using Thermalherd to control processors.

“ The distributed nature of Thermalherd lends itself to the policing of multiple processors as well,” she said. “So the policemen at each junction would not just regulate network traffic. They’d also drive each processor that was stuck in that traffic.

“ It’s exciting,” she said. “It’s hard but it’s fun.”



Benziger asks if fuel cells are really wave of the future

Environmentally aware citizens are being spotted more often driving hybrid vehicles, energy-efficient alternatives to the standard internal combustion vehicles. As hybrid vehicles become old hat, the buzz on fuel cells—a possible alternative to the internal combustion engine—is still fresh.

A fuel cell is an electrochemical energy-conversion device; it converts hydrogen and oxygen into water, producing electricity in the process. It produces less pollution than fossil fuels.

“ If you looked at the sales pitch, they are a nonpolluting, fuel-efficient power method,” said Jay Benziger, professor of chemical engineering. “But there are problems with fuel cells. They don’t function nearly as well as claimed.”

Professor Benziger and his team inspect these problems, and hunt for solutions. They work with the chemistry department on new materials for fuel cells, experiment with some of the engineering concepts that could improve fuel cells, and assess their utility.

How they work
Fuel cells create direct current (DC) voltage through an electrochemical reaction. There are several different types of fuel cells, classified by the type of electrolyte they use in the reaction. The proton exchange membrane (PEM) fuel cell is the type that researchers are looking at to power vehicles and homes.

Oxygen and hydrogen are forced into the fuel cell. The gases are then distributed over the surface of the membrane, which is coated with a catalyst, usually made of plati-num powder.

The gases react to create water and electrical energy.

In theory, fuel cells are simple, but Professor Benziger and his research team have found that in practice they are quite complicated.

Photo by Frank Wojciechowski Jay Benziger conducts research on a variety of energy technologies.

Hydrogen issue
“ One of the biggest problems is what hydrogen you will get to run your fuel cells,” said Professor Benziger. “Oxygen comes from the air, but hydrogen does not occur naturally. So where do you get it from?

“ Well, the method now is you get it from fossil fuels.”

When fossil fuels react with steam, they are converted into carbon dioxide, carbon monoxide (CO), and hydrogen. Through further processing, the hydrogen can be partially purified, but not completely.

“ Under the conditions, the best you can do is to get down to about 1,000 parts per million of CO in your hydrogen,” Professor Benziger said. “In these PEM fuel cells, what happens is the CO competes with hydrogen to absorb on the electrode surface. At the temperatures they would like to operate at, the electrode gets completely covered with CO, makes no contact with hydrogen, and so the fuel cell doesn’t work.”

Experimentation showed Professor Benziger and his team that the battle for the electrode could be tipped in hydrogen’s favor by raising the temperature at which the reaction occurs to about 130 degrees Celsius.

There are two problems, however, with raising the reaction temperature that high. First, the chemists would have to find membrane materials that could function at that temperature. Second, membranes need to contain a constant, sufficient supply of water within them to conduct the protons that allow the energy current to flow. At normal atmospheric pressure, however, water vaporizes at only 100 degrees Celsius, drying out the fuel cell and halting the reaction that makes it work.

Water issue
A discovery made by Professor Benziger’s team is that there is a critical amount of water needed to start a fuel cell, and, further, a critical amount to keep it running.

“ If the fuel cell doesn’t have enough water, the gases flowing through it dry out the fuel cell, and it doesn’t work,” said Professor Benziger. “With enough water, the reaction proceeds fast enough to continue making enough water to keep it going. We call this the ‘ignition condition.’ We titled our paper on it, ‘You can fan the flame with water.’”

At a certain controlled pressure, water’s boiling point could be raised high enough to create this ignition condition. However, use of a fuel cell in most of its proposed applications, like running a car, will not occur in very controlled conditions.

“ This is a key engineering concept,” Professor Benziger said. “Real use of the fuel cell will not be like putting it in a laboratory where you run it in controlled temperatures and pressures at steady state all the time. I want to know that if I’m in Northern Minnesota in the middle of winter or the Arizona desert in the middle of summer, I can be sure that the fuel cell will start up.”

While working on the real-life engineering of the fuel cell’s water supply, Professor Benziger and his team also began considering other challenges to the fuel cell operation in real-life applications.

Variable load issue
Driving a car uphill against the wind from a dead stop requires a car engine to work differently than it will when the car is cruising down a highway. In this and many other instances, a fuel cell would need to be able to handle variable loads.

Photo by Frank Wojciechowski Jay Benziger, professor of chemical engineering, researches the engineering of fuel cells.

“ The thing is, we’ve discovered that there are quite complex dynamics if you change the load on a fuel cell,” Professor Benziger said. “Previous work had focused on just keeping things running at steady state. But if you’re running a car or doing various other applications, you’ll have variable loads.

“ At variable loads, because you make water in the fuel cell, because it makes more or less water depending upon the load, and because the water affects the operation of the fuel cell,” he said, “the dynamics of the operation change as the load changes.”

Even at steady, fixed, controlled conditions, the researchers discovered that the fuel cell’s output, over a period of several hours, would vary by a factor of two.

“ This comes about, we believe, because the membranes swell and compress due to the water going in and out,” said Professor Benziger. “This gives rise to very unusual behavior.”

Complexity issue
Professor Benziger said that these complexities make development of proper fuel cell systematics and operation of fuel cells difficult.

“ Because different parts of a fuel cell act out of phase with each other, you just get massive chaos,” he said. “As I put it, I’m not smart enough to understand that.”

Professor Benziger and his group have been developing a simplified, one-dimensional version of a fuel cell, which has thus far provided “very clean, regular effects.”

Is it necessary?
“ Of course, one of the other problems is whether or not fuel cells even make sense for these applications,” Professor Benziger said. “Even though we’re doing a lot of interesting research, my opinion is that a fuel cell really does not make sense for a car.”

He explained that although a hydrogen-powered fuel cell itself may reasonably claim to be 50-percent efficient—which is superior to the 25-percent efficiency of the average internal combustion engine—a complete well-to-wheel assessment tells a different story.

The hydrogen needed to power the fuel cell is created by the consumption of fossil fuels. When this is factored in, the overall efficiency of a fuel cell is about the same as the internal combustion engine.

“ And it’s more complicated to operate,” Professor Benziger said.

In addition, the fuel cell does not provide freedom from the nonrenewable fossil fuels that proponents of sustainability yearn for.

Feasible uses
Nonetheless, Professor Benziger believes that fuel cells have their place in the energy-efficient future.

“ Where I personally think you’ll see fuel cells first making inroads are niche markets where convenience is more important that economics,” he said.

Professor Benziger said that the lightweight fuel cell could replace heavy batteries for laptop computers or in special military applications.

He also believes that fuel cells could make their long-term impact by doing load leveling for renewable power generators like solar panels and windmills.

“ The sun doesn’t shine at night and the wind doesn’t blow all the time, so you need to do load leveling,” he said. “A way to do this is, when you’re getting the most amount of energy from the solar panel or windmill, you divert part of that to run an electrolysis cell to generate hydrogen. Later on, when the solar or wind unit is not producing, you take that hydrogen, run it through a fuel cell, and run a constant load.”

Profesor Benziger said that his skepticism about fuel cells’ ability to power cars causes mixed reactions.

“ There have been some people who are very excited about our work,” he said, “and some people who don’t like it at all, because it upsets the way they think.”



EQuad News peeks inside more energy labs

Developing low-emissions power sources is a main goal of the Carbon Mitigation Initiative at Princeton. C.K. Ed Law, professor of mechanical and aerospace engineering (MAE) and MAE Assistant Professor Yiguang Ju are studying the combustion of alternative fuels, including hydrogen, biomass, and synthetic fuels. Professor Law studies hydrogen fuel, which has several inherent challenges. Hydrogen is not a naturally occurring gas, storage is tricky due to the hydrogen’s high flammability and combustibility, and hydrogen must be supercharged to provide enough power for most needs. The supercharging process damages fuel efficiency. Professor Law and his group seek solutions to hydrogen’s combustion and safety problems. Research has thus far shown that mixing propane with hydrogen can help satisfy both of these issues, although more research is needed. Professor Ju’s group is studying a synthetic fuel called dimethyl ether. The group hopes to create successful combustion models for numerical prediction and extend the current research for industrial applications.

Professor of Operations Research and Financial Engineering René Carmona studies the importance of risk management at all points in the production and delivery chain of energy markets. He conducts statistical analysis of energy data and develops mathematical models for energy pricing. Professor Carmona also contributes to discussion of energy markets through his teaching and professional activities. He teaches ORF 569: Risk Management for the Energy Market and recently co-organized a conference titled “Price Risk and the Future of Energy Markets.” The conference focused largely on the extensive debate over deregulation of electric utilities (www.princeton.edu/~seasweb/eqnews/winter03-04/feature3.html).

Researchers in the lab of Margaret Martonosi, professor of electrical engineering, are working with researchers in the Department of Ecology and Evolutionary Biology on a project that is creating many novel approaches to advance energy-aware computing. The project, dubbed Zebranet, is a wireless sensor network being established in the Mpala Research Centre in Kenya to track and collect data about zebra herds. The sensor network must be hardy and largely self-sustaining, but also energy-efficient. The system must be very aware of weight, since each node of the network will be in the form of a collar worn by a wild zebra. However, the node must contain a global positioning system, storage cells, a wireless transceiver, and a CPU, leaving little room for a power source. Batteries are very heavy, so the engineers are creating a power supply system, in which the battery will be able to operate for five full days between recharges by solar panels that line the collar. For more information, see www.princeton.edu/~seasweb/eqnews/spring03/feature4.html.

Researchers in the lab of Stephen Forrest, professor of electrical engineering, have invented a technique for making organic solar cells that, when combined with recent advances, could yield a highly economical source of energy. Solar cells convert light to electricity and are already used to power many devices, from calculators to satellites. Conventional solar cells operate at approximately 24-percent efficiency, meaning they convert 24 percent of the available light energy into electrical energy. However, these conventional solar cells, while highly efficient, are made with expensive, silicon-chip-based technology, limiting their usefulness in the marketplace. Conversely, organic solar cells had, until recently, been far cheaper to manufacture, but were only 1-percent efficient. By changing the combination of organic compounds normally used to make organic solar cells, Professor Forrest’s group has created cells are just over 3-percent efficienct. The researchers are confident they can soon increase efficiency to 5 percent by combining their new materials with manufacturing techniques that were recently advanced. They believe that low-cost solar devices with 5-to 10-percent efficiency could be viable in the marketplace.

The Princeton Plasma Physics Laboratory, with which SEAS faculty are affiliated, will be the United States project center of ITER, a major international magnetic fusion experiment conducted by the U.S. Department of Energy. The purpose of ITER is to test the feasibility of nuclear fusion as a source of electricity and hydrogen. Proponents of fusion power say that such a power plant would produce no greenhouse gas emissions, use abundant and widely distributed sources of fuel, shut down easily, require no fissionable materials, and produce manageable radioactive waste.

The pulp and paper industry is among the largest producers and users of renewable energy in the United States today. The industry creates and consumes energy from woody biomass and “black liquor,” a pulping-process residue remaining after removal of the cellulose for papermaking. Presently, black liquor is consumed in liquid form by conventional burners, but gasification of the black liquor could greatly increase its efficiency. With such renewable energy resources at its disposal, there is significant potential for this industry to catalyze the development of a “biorefining” industry that would generate substantial amounts of renewable energy. The Princeton Environmental Institute (PEI), with ehich SEAS faculty are affiliated, is conducting a cost-benefit analysis of this potential, studying its impacts upon energy savings, energy security, rural development, and the environment. The lab of Eric Larson, PEI research engineer, will conduct the study. The study is one of 22 projects being funded by the Biomass Research and Development Initiative, led by the U.S.Departments of Energy and Agriculture.

 

[ contents ]   [ previous story ]  [ next story ]   [ top of page ]