Earticle

초록

영어

The proliferation of multimodal systems demands efficient management of heterogeneous computing resources. However, most GPU-centric frameworks still rely on static scheduling, resulting in unbalanced utilization and energy waste. This paper presents HERMES (Heterogeneous Efficient Resource Management and Execution Scheduling), an adaptive scheduling framework designed for efficient scheduling in heterogeneous multimodal AI systems. HERMES introduces HScore, a unified metric that quantifies heterogeneous efficiency by integrating performance (FPS) and power consumption. Experimental results on a ViT-based multimodal benchmark show that HERMES achieves up to 12.7% faster execution and 15.8% higher energy efficiency than static hybrid baselines, while maintaining balanced CPU–GPU utilization. These findings confirm that adaptive feedback scheduling significantly enhances both scalability and sustainability in multimodal AI systems.

목차

Abstract
I. INTRODUCTION
II. METHODOLOGY
A. Overall Framework
B. Definition of Efficiency Metric (H-Score)
C. Algorithm Design
D. Summary
III. EXPERIMENTAL RESULTS AND DISCUSSION
A. Experimental Environment
B. Evaluation Metrics
C. Results and Analysis
D. Discussion
IV. CONCLUSION AND FUTURE WORK
REFERENCES

저자

• Yujin Ju [ Department of Artificial Intelligence Convergence, Graduate School Dankook University Yongin-si, Korea ]
• Sungshin Kwak [ Department of Artificial Intelligence Convergence, Graduate School Dankook University Yongin-si, Korea ]
• Jihyeok Park [ Department of Softwre Science, Graduate School Dankook University Yongin-si, Korea ]
• Jaihee Cho [ Department of Data and Knowledge Service Engineering, Graduate School Dankook University Yongin-si, Korea ]
• Seongje Cho [ Department of Software Science Dankook University Yongin-si, Korea ]
• Eungkyo Suh [ Department of Business Administration Dankook University Yongin-si, Korea ]
• Sohyun Park [ Department of Software Science Dankook University Yongin-si, Korea ] Corresponding Author

참고문헌