Electrical and computer engineering (ECE) Ph.D. graduate Nicholas Woolsey, Associate Professor Rong-Rong Chen and Assistant Professor Mingyue Ji won the Best Paper of the Communications Theory Symposium at the 2021 IEEE Global Communications Conference (GLOBECOM). They co-authored the paper with New Jersey Institute of Technology ECE Professor Joerg Kliewer. Their paper, “A Practical Algorithm Design and Evaluation for Heterogeneous Elastic Computing with Stragglers,” introduces a novel algorithm that enables systems to handle virtual machines with different and/or slower computing speeds. The algorithm minimizes computation time for modern cloud computing platforms such as Amazon Web Services, Microsoft Azure and Google Cloud Platform. The work was part of Woolsey’s Ph.D. dissertation, which won the 2021 Best ECE Dissertation Award.
“It’s exciting to have our paper acknowledged as a valuable part of the research happening around the world and to be recognized by the community for our hard work,” Woolsey said. “This research can be applied to so many different applications in distributed computing and machine learning as well as reduce the cost for academia and industry using computing resources.”
Distributed computer systems consist of several software components spread across multiple computers but run as a single system. It allows large amounts of data that would otherwise not fit on a single machine to run efficiently and reliably across multiple. Many companies today rely on distributed systems like cloud computing to quickly increase or decrease their computer processing, memory, and storage needs based on demand. There are, however, limitations to these types of systems, including privacy concerns, computation and storage limitations, and machine failures. Current research utilizes code to resolve many of these issues but generally assumes that all machines have the same computation speed.
The group’s work demonstrates that often this is not the case, and frequently machines have different or heterogeneous computation speeds even if they share the same configuration. Their paper outlines a new framework that can optimally handle heterogeneous computation speeds across machines and introduces straggler tolerance using a novel algorithm. Based on evaluation of the performance of their designs, the group found that the algorithm outperforms state-of-the-art design by more than 30%.
The Communication Theory Symposium is one of 12 symposia included in the GLOBECOM conference and focuses on the fundamental and theoretical aspects of communication systems with an emphasis on wireless and wireline communications. GLOBECOM is one of the IEEE Communications Society’s two flagship conferences dedicated to driving innovation in nearly every aspect of communications.