Park Junghyun and Kim Daero of Dong-A University Present Paper at Top Cloud Computing Conference

'IEEE International Conference on Cloud Computing 2024' Held in Shenzhen, China

Dong-A University (President Lee Haewoo) announced that master's students Park Junghyun and Kim Daero from the Department of Computer Engineering presented an outstanding paper at the 'IEEE International Conference on Cloud Computing (CLOUD) 2024,' which is recognized as the premier academic conference in the field of cloud computing.

This conference, recently held in Shenzhen, China, is also designated as an excellent academic conference by the National Research Foundation of Korea (NRF).

The title of the paper presented by the Dong-A University research team is 'Carbon-Aware and Fault-Tolerant Migration of Deep Learning Workloads in the Geo-distributed Cloud.'

This paper, supervised by Professors Han Jungkyu and Chun Sejin of the Department of Computer and AI Engineering at Dong-A University, lists master's students Park and Kim as the first authors, with Kim Jiseon, a student from the Department of AI, as a contributing author.

In this paper, the research team addressed the issue of reducing carbon emissions, which is a major topic worldwide and a key issue in Busan. AI training is mainly conducted in cloud environments, and large language models (LLMs) such as OpenAI's ChatGPT consume massive amounts of electricity and emit carbon. In particular, carbon emissions from power generation for data centers have been continuously increasing every year.

To solve this problem, the Dong-A University research team recognized that cloud data centers are distributed globally and that the associated power grids have varying carbon emissions over time. Based on these characteristics, they proposed a migration method that transfers deep learning model training to cloud centers where power usage and expected carbon emissions are minimized.

Previous studies mainly focused on reducing either power consumption or carbon emissions in fixed regional clouds. However, this study is significant in that it proposes a core technology for cloud middleware that accommodates the characteristics of distributed data centers and temporary power peaks.

This approach can effectively minimize carbon emissions by distributing services to data centers according to time and space, as carbon emissions differ by domain. It is expected to make a technical contribution to measuring and reducing digital carbon emissions for various web hosting and app services of public institutions and companies deployed in the cloud.

Park said, "I am delighted that the hard work we put in day and night at the lab has led to such a valuable achievement," and added, "I hope to become a talent who can contribute to the core technology for reducing carbon emissions from data centers in Busan in the future."

Kim stated, "Through this international conference, I was able to interact with outstanding researchers from around the world and realize that the issue of reducing digital carbon emissions is rapidly emerging as a major global issue," and added, "It is meaningful that the proposed technology can be applied as a core technology in cloud middleware."

Professor Chun said, "Although sustainable computing research to reduce digital carbon emissions, including in the cloud, is very active overseas, only a few researchers in domestic institutions and companies are conducting such studies," and added, "I believe that systematic collaboration and research organizations for technology development in the field of digital carbon emissions are needed in Korea as well."