Prof. Javad Ghaderi Wins NSF CAREER Award for Network Optimization
Javad Ghaderi, assistant professor of electrical engineering, has won a National Science Foundation CAREER award for his work on a new algorithmic framework for optimizing the performance of large-scale networks, including wireless network systems and data centers. His research could help improve emerging applications in areas such as disaster recovery, healthcare, tracking, data processing, and computing, as well as the use of wireless networks in our daily lives.
“As wireless network systems and data centers continue to scale up, optimization problems have become even more complex,” says Ghaderi, who is also a member of the Data Science Institute. “This is especially true now in the age of the Internet of Things and big data, where we are witnessing the expansion of networks at large scale which require highly scalable and efficient resource allocation algorithms.”
Our modern world of expanding communication networks faces many challenges in optimizing resources: everyone wants information to be available immediately, with minimal delay, minimal energy consumption, and no bumps in the flow of data. Ghaderi is tackling optimization challenges that arise in large networks by examining ways to exploit their stochasticity (randomness) and large-scale nature to design simple adaptive algorithms that can efficiently solve the optimization problems.
Communication networks are inherently stochastic. The times when data packets arrive at a wireless device vary randomly, as do the times when data processing jobs are submitted to a data center, as do the processing requirements of the jobs. Most of the resource optimization problems in these networks require solving hard combinatorial problems: when and how to distribute and execute the data processing jobs, managing the data flows in data centers, and when and how to transmit data packets in a wireless system in order to maximize throughput, minimize delay, or minimize energy consumption.
Ghaderi’s team is focused on setting the foundation of this new era of algorithms for such large-scale stochastic networks through an interdisciplinary approach that unifies combinatorial and stochastic optimization. His research has shown that, somewhat counter-intuitively, system stochasticity coupled with large scale might actually help in designing simple, low-complexity algorithms that can approach optimal solutions over time, or as networks grow. The impact of “wrong temporal/spatial decisions” made by simple algorithms vanishes as the network dynamics continue over time or the network size scales up, Ghaderi explains. “Thus, simple randomized algorithms can provide performance guarantees for problems that are otherwise very hard to solve,” he says.
Ghaderi is also collaborating with several colleagues at Columbia Engineering to implement his algorithms in practice. The researchers are looking specifically at two problems: (1) implementing the algorithms for flow scheduling in data centers using CloudLab, an NSF-funded cloud platform for hosting experimental prototypes; (2) implementing algorithms for wireless scheduling on a testbed of energy-harvesting devices at Columbia Engineering. His work on scheduling algorithms for ultra-low-power wireless devices recently won the Best Paper Award at ACM CoNEXT 2016, a major conference for novel networking technologies.
Ghaderi’s CAREER project is aimed at inventing a general mathematical framework that goes beyond wireless and data center networks and can be used in many modern applications that require combinatorial decision making in the presence of uncertainty. “I am very interested in the analysis, design, and management of large-scale networked systems,” Ghaderi says. “I think this new project will create a new way of solving optimizations in modern systems, enabling us to allocate resources much more effectively.”
The five-year, $500,000 award, part of the NSF’s Faculty Early Career Development (CAREER) Program, is one of the NSF’s most prestigious honors for faculty at or near the beginning of their careers.
—by Holly Evarts