SecuLayer Announces Platform Technology That Delivers the Performance of 15 Servers with Just One at 'STICK 2022'

SecuLayer held the developer conference 'STICK 2022', where researcher Baek Man-ki presented a platform technology that achieves the performance of 15 servers using just one machine learning server.

During the session titled 'Creating One Server Equivalent to 15 Machine Learning Servers! GPU Resource Allocation with eyeCloudAI', held on the 21st at the Seoul Forest Galleria Foret, researcher Baek introduced real-world examples of SecuLayer's platform technology implementation.

Researcher Baek explained, “By allocating GPU memory in blocks, we improved GPU utilization efficiency. Furthermore, using GPU acceleration technology, we implemented a method that processes models approximately 56 times faster than before while maintaining high accuracy.”

He also emphasized the importance of platform technology that effectively distributes GPU resources. Baek stated, “At a client company, we improved deployment and management convenience based on Kubernetes, achieving the effect of 15 servers with just one GPU. When converted to server savings, this amounts to a total of 336 servers, resulting in a cost-saving effect of approximately 27.55 billion won.”

He concluded by saying, “SecuLayer applied GPU resource allocation technology to our AI platform 'eyeCloudAI', enhancing our clients' security self-sufficiency while driving time and cost savings. We will continue to focus on research and development to implement a platform that satisfies both technological and economic aspects.”