UNIVERSITY PARK, Pa. -- Many iPhone users are familiar with using iCloud, a virtual storage device that saves data such as photos, music and other information from your device. But for computer science and engineering graduate students Cheng Wang and Neda Nasiriani, cloud computing became the focus of a research paper that landed their team the runner up for best paper at this year’s prestigious ACM International Conference on Future Energy Systems.
The conference, which was held in July in Bangalore, India, is “the premier venue for researchers working in the broad areas of computing and communication for smart energy systems, and in energy-efficient computing and communication systems,” according to the conference’s website.
Over 70 organizations and large companies throughout the world such as IBM submitted papers to the conference, but only 16 were accepted. Wang and Nasiriani’s team’s 10-page research, “Recouping Energy Costs from Cloud Tenants”, was one of those few that made it through several rounds of judging, rebuttal and presentation, eventually leading to their award of runner-up for best paper.
“We were notified on the acceptance of the paper on April 1 and I remember thinking, ‘Is this an April Fool’s joke?’” Wang said.
After being approached by their Penn State advisers George Kesidis, Bhuvan Urgaonkar and Qian Wang to submit a paper for the conference, Wang and Nasiriani decided to collaborate with them and formed a research group in October 2014.
The idea for their research, which focused on how cloud providers can design a dynamic pricing scheme for charging its customers to improve profitability, stemmed from Wang’s pervious internship at IBM.
To obtain and apply real-world numbers on the profitability of the cloud at a large company, the group added Lydia Chen and Robert Birke of IBM Research Zurich, Switzerland and Aayush Gupta of IBM Research Almaden in San Jose, California.
The group found that the cloud charges customers a flat rate per hour of use, but the electricity bill of the cloud, which accounts for 10 to 20 percent of the cloud’s overall cost, is time varying.
Wang said for example in the daytime, a cloud’s electricity price may be very high and in the evening, it could be significantly lower. Because of this, the energy cost of the cloud is time varying, yet the cloud charges its customers a standard flat rate.
The group decided a more dynamic pricing scheme for charging customers would be the answer to improving the cloud’s profitability by 30 percent. “When the electricity price is high, you can charge the customers a higher price to help generate more profit,” Wang said.
Together, the group designed their own frameworks and algorithms to deal with complexities such as customer demand and utility functions. They came to the conclusion that at the beginning of each hour, the cloud should set a price based on their algorithm and then in the same hour the customer has to react to this price, creating a leader and follower game.
After working 40 hours a week for several months and coordinating dozens of in-person meetings and Skype calls, their research culminated at the conference. Though the group could not attend in person, Wang said he did a video presentation and phone call for the panel of judges who named them runner up for best paper.
Nasiriani said getting to practice the theories she has learned throughout the years while bringing new insights into the electrical engineering environment was the most beneficial aspect of participating in the conference.
Wang said he was grateful to learn mathematical techniques that he was not familiar with before, but he is most proud of the work that he and his team were able to accomplish.
“We have a good paper, we have solid work and people are starting to recognize our work,” he said.